Dear {{first_name | reader}},

This month closes out an AI-driven protein robustness study focused on what happens to experimentally validated, AI-designed protein binders under structural stress.

What started as a technical investigation into mutation responses became something more fundamental: a design question about the difference between working once and actually being reliable.

Between experimental validation and reliability

The current standard of success in AI protein design is essentially: can we produce a protein that performs the intended function? Increasingly, yes. AI-designed proteins now bind to targets with high precision, behave as expected in lab experiments, and pass experimental validation. At that point, the design is usually considered done.

But this project began from a different assumption: that working once doesn't necessarily mean something is reliable. Most AI-driven protein design systems are optimised to answer that first question and stop there.

Structural predictions, confidence scores, and lab validation tell you that something works in a specific setup. They don't tell you how stable the structure is under change, where the weak points are, or whether the design can be trusted beyond that initial success.

The study took AI-designed binders that had already passed experimental validation and asked what happens when you push them, introducing small controlled changes rather than designing new proteins from scratch.

Three patterns came up consistently.

1. Working can hide fragility: some proteins look structurally clean, score well, and pass validation, but destabilise under small changes.

2. Reliability is uneven: within the same design system, some proteins remain stable under perturbation while others sit close to failure.

3. Failure is often invisible: in several cases the overall structure stays intact, but the functional region where binding happens fails. The artifact looks stable but has lost its intended behaviour.

One further finding worth noting: attempts to benchmark these proteins against natural equivalents often failed, not due to poor matching but because there simply isn't a clear natural equivalent.

AI-designed proteins may occupy regions of design space biology never explored, which means reliability has to be actively constructed through evaluation rather than inferred from nature.

What this looks like in practice: bacterial cellulose and living materials

Bacterial cellulose is one of the more compelling materials in biodesign right now: a nanofibrous scaffold produced by bacteria, mechanically strong, highly pure, and finding applications in wearable textiles, wound dressings, and flexible electronics.

Controlling its properties precisely remains a significant challenge, and this is where AI-designed proteins are starting to enter the picture. Researchers are exploring how engineered proteins can direct cellulose assembly, modify fibril surface chemistry, or interface the material with biological and synthetic components.

But bacterial cellulose fabrication is not a controlled laboratory environment. The bacteria are subject to temperature variation, nutrient fluctuation, and mechanical agitation across days or weeks of production.

A protein that functions reliably in an optimised lab setup may behave very differently inside a living production system. If that protein sits close to a structural failure threshold, as our study suggests many AI-designed proteins do, those fluctuations may be enough to push it over quietly, in ways that don't surface until much later in the pipeline.

The same logic extends to engineered silk, mycelium composites, and biofilm-based structures. Anywhere the designed biology has to perform reliably across a production process rather than a single validated experiment, structural robustness under variation is not a secondary concern. It is central to whether the design actually works.

Toward trust-aware biodesign

The paper formalises this into a post-success evaluation framework: a structured method for stress-testing protein designs and distinguishing hidden fragility from genuine robustness. In traditional design disciplines, we don't stop at "it works."

We ask how something behaves under stress, where it fails, whether it holds across conditions. In AI-driven biodesign, we're mostly still at "it works, move on." The argument here is for a different posture: it works, now evaluate whether it holds.

That shift matters as AI-designed proteins move closer to deployment in therapeutics, biosensors, molecular control systems, and living material fabrication. Treating experimental validation as the finish line made sense when getting proteins to work at all was the hard part. That's no longer the hard part.

The full paper is coming soon. If any of this connects with work you're doing in protein design, living materials, or the broader question of how we build trustworthy biological systems, I'd be glad to hear from it.

Until next time,

Raphael

What Happens After AI-Designed Proteins “Work”? A Framework for Reliability, Robustness, and Trust in Biodesign

Key Insight

AI-designed proteins can pass experimental validation and still fail under real-world conditions. True success in protein design is not whether a system works once—but whether it remains stable, functional, and reliable under variation.

Why “It Works” Is No Longer Enough in AI Protein Design

The current benchmark in AI-driven protein design is functional success:

• Does the protein bind its target?

• Does it behave as expected in a controlled experiment?

Increasingly, the answer is yes.

However, experimental validation only confirms performance under a specific set of conditions. It does not evaluate:

• Structural stability under perturbation

• Sensitivity to mutation or environmental change

• Reliability across time, scale, or production environments

This creates a critical gap between initial success and real-world reliability.

From Experimental Validation to Structural Reliability

This study began with a different assumption:
A protein that works once is not necessarily a protein that can be trusted.

Methodology Overview

• Selected AI-designed protein binders already experimentally validated

• Introduced small, controlled perturbations (e.g., mutations)

• Observed structural and functional responses

Rather than designing new proteins, the focus was on stress-testing existing designs.

Three Core Findings from Protein Stress Testing

1. Functional Success Can Mask Structural Fragility

• Proteins with high confidence scores and clean structures may still destabilize under minor changes

• Standard evaluation metrics often fail to detect hidden weaknesses

2. Reliability Is Uneven Within the Same Design System

• Some proteins remain stable across perturbations

• Others operate close to structural failure thresholds

This suggests that robustness is not guaranteed—even within high-performing AI pipelines.

3. Failure Is Often Invisible

• Structural integrity may appear intact

• Functional regions (e.g., binding interfaces) can fail silently

Result:
A protein may look correct but lose its intended behavior.

Why Natural Benchmarks Don’t Always Apply

Attempts to compare AI-designed proteins with natural equivalents revealed a limitation:

• Many designs do not map to known biological structures

• These proteins may occupy novel regions of design space

Implication

Reliability cannot be inferred from nature.
It must be explicitly engineered and evaluated.

Real-World Application: Bacterial Cellulose and Living Materials

What Is Bacterial Cellulose?

• A nanofibrous material produced by bacteria

• Known for strength, purity, and flexibility

Current Applications

• Wearable textiles

• Wound dressings

• Flexible electronics

The Role of AI-Designed Proteins

Engineered proteins are being explored to:

• Direct cellulose assembly

• Modify fibril surface chemistry

• Enable integration with biological and synthetic systems

The Reliability Problem in Living Systems

Unlike controlled lab environments, biological production systems introduce variability:

• Temperature fluctuations

• Nutrient variability

• Mechanical stress over time

Critical Risk

Proteins designed near failure thresholds may:

• Degrade under production conditions

• Lose function without visible structural failure

• Fail late in the development pipeline

Broader Relevance

This challenge extends to:

• Engineered silk

• Mycelium composites

• Biofilm-based materials

In all cases, robustness under variation determines success.

Toward a Post-Success Evaluation Framework

What Is a Trust-Aware Biodesign Approach?

A shift from:

• “It works → move on”

To:

• “It works → now test whether it holds”

Core Components of the Framework

EEAT in Practice: Experience, Expertise, and Evidence

First-Hand Research Experience

This framework is grounded in direct experimental analysis of AI-designed proteins under controlled perturbations, rather than theoretical modeling alone.

Alignment with Emerging Research

Recent advances in protein design (e.g., deep learning-based structure prediction and generative models) have significantly improved functional success rates. However:

• Studies show mutation sensitivity remains a major limitation

• Protein stability remains a key bottleneck in therapeutic and industrial deployment

Multiple Perspectives

• Engineering view: Success = function achieved

• Biological systems view: Success = function sustained under variability

• Design perspective: Success = reliability across contexts

Limitations and Open Questions

• Stress-testing increases evaluation complexity and cost

• No universal benchmark for robustness currently exists

• Trade-offs between performance and stability remain unresolved

Why This Matters for the Future of Biodesign

As AI-designed proteins move toward deployment in:

• Therapeutics

• Biosensors

• Molecular control systems

• Living materials

The definition of success must evolve.

Key Shift

• Past: Difficulty was making proteins work

• Present: Difficulty is ensuring they remain reliable

Summary

• Experimental validation confirms function—but not reliability

• AI-designed proteins often exhibit hidden fragility

• Robustness must be actively tested, not assumed

• Living systems amplify the consequences of instability

• A post-success evaluation framework is essential for trust in biodesign

FAQs About Protein Reliability in AI-Driven Design

What is the difference between validation and reliability?

Validation confirms a protein works under specific conditions. Reliability ensures it continues to work under variation, stress, and real-world environments.

Why do AI-designed proteins fail after initial success?

They may be structurally fragile or sensitive to small perturbations not captured during initial testing.

Can natural proteins be used as benchmarks?

Not always. Many AI-designed proteins exist in novel design spaces without natural equivalents.

What industries are most affected by protein reliability?

• Biotechnology and therapeutics

• Materials science (e.g., living materials)

• Synthetic biology and biofabrication

How can reliability be improved?

Through systematic stress testing, mutation analysis, and evaluation across environmental conditions.

Final Takeaway

AI has solved the problem of making proteins work.
The next challenge is ensuring they keep working.

Reliability is not a byproduct of design—it is a requirement that must be engineered, tested, and validated explicitly.

Dear {{first_name | reader}},

What does it mean to trust a molecule?

That question sits at the center of our new publication this month, on Trusted Molecular Memory (TMM), a framework for thinking about trust in systems where biology, computation, and institutional processes all meet.

The basic idea is simple: a molecule can carry evidence, but it cannot prove trust by itself. Trust depends on the full context around that evidence, including how it was made, handled, measured, interpreted, and governed.

Why this matters

In molecular systems, the data can look correct and still leave important questions unanswered. Where did the sample come from? Has it been changed? Was it measured in a way that altered it? Who gets to decide what counts as a valid result? TMM argues that these are not side questions. They are part of trust itself.

From verification to trust

TMM brings together five molecular trust pillars: 1) Material Evidence, 2) Process & Provenance, 3) Interpretation & Confidence, 4) Accountability & Contestability, and 5) Governance. It also introduces a Molecular Chain of Custody (M-CoC), a way to track how a molecular artefact is produced, stored, measured, transferred, regenerated, and interpreted over time.

The key point is that verification and trust are not the same thing. A result can be verified and still remain uncertain in context.

To make TMM more concrete, we developed a set of scenarios that follow molecular artefacts through real-world-style situations. We wanted to stress-test the framework in cases where the molecule itself may still be intact, but trust becomes uncertain because of regeneration, repeated measurement, or institutional change.

These scenarios are not predictions or edge cases for their own sake. They are design probes that help show where trust can hold, where it can weaken, and where a technically correct result may still be hard to trust in practice.

Scenario A: Regeneration and material continuity

In the first scenario, a voice message is encoded into DNA, verified, and sealed away for long-term storage. Years later, someone wants to hear the message, but opening the sealed sample would consume part of the material. So the system regenerates a fresh DNA instance from the stored digital encoding and decodes that instead. The recovered audio is the same, but the physical molecule is not the same one that was originally synthesized.

This scenario shows that informational continuity and material continuity can diverge. The message remains intact even when the original molecule no longer exists in the same form.

Scenario B: Reading as intervention

In the second scenario, a museum stores historical letters in DNA and uses sequencing to check that the archive remains readable. The problem is that every read changes the material a little, because molecular measurement is not fully passive. Over time, repeated checks and access copies slowly reshape the lineage of the archive, even though the encoded text remains recoverable.

This scenario shows that verification itself can become a custody event. In other words, the act of reading can alter what is being trusted.

Scenario C: Institutional discontinuity

In the third scenario, a legal will is encoded into DNA and stored in a licensed archive under notary supervision. Years later, the archive is reorganized, systems are migrated, and staff change, but the capsule itself remains sealed and chemically stable. When the DNA is eventually opened in court, the decoded text matches the original perfectly. Even so, the dispute is not about the molecule. It is about whether the chain of custody and institutional authority remained continuous.

This scenario shows that material stability does not guarantee institutional trust.

What these scenarios reveal

Taken together, the scenarios show a new failure class: valid data, wrong decision. A molecular result can be correct, while the broader context still makes the conclusion uncertain. TMM shifts the focus from asking whether a molecule is authentic in isolation to asking how trust emerges across evidence, process, interpretation, and governance over time.

Why AI makes this more urgent

This becomes especially important as AI agents begin to design sequences, interpret outputs, and help make decisions in biological workflows. AI can be very good at classifying molecular data, but molecular data is not ground truth in a vacuum. It is context-dependent and historically situated.

A system can act on valid data and still make an untrustworthy decision if it does not know how that data was produced, handled, or transformed. TMM offers a way to build trust-aware bio-digital systems that reason about consistency, provenance, and institutional context, not just correctness.

Why biodesigners should care

For anyone working with DNA storage, living materials, or bio-digital interfaces, the lesson is that change is not always noise. In these systems, change may be part of the evidence itself. TMM gives us a language for designing with that reality instead of pretending molecules behave like static digital files.

Read the full paper here: https://doi.org/10.5281/zenodo.19002474

This work was developed with Dr Larissa Pschetz (University of Edinburgh), Joe Revans (Fallow Earth), and Prof. Thomas Heinis (Imperial College), and funded by the Advanced Research and Invention Agency through the Trust Everything Everywhere programme.

Until next time,

Raphael

Founder, Biodesign Academy

What Is Trusted Molecular Memory (TMM) and Why Biology Cannot Be Treated Like a Digital File

Introduction: What Does It Mean to Trust a Molecule?

Trusted Molecular Memory (TMM) is a framework for understanding trust in systems where biology, computation, and institutional processes intersect. The core principle is simple but critical:

A molecule can carry evidence, but it cannot establish trust on its own.

Trust emerges from the full lifecycle context—how molecular data is created, handled, measured, interpreted, and governed.

This article explains how TMM reframes trust in molecular systems, why verification is insufficient, and how this impacts AI-driven biodesign.

Why Trust in Molecular Systems Is More Complex Than It Appears

Molecular data can appear technically correct while still raising unresolved questions:

• Where did the sample originate?

• Has the material been altered or regenerated?

• Did measurement processes change the sample?

• Who defines what counts as a valid result?

Key Insight:
These are not peripheral concerns—they are core components of trust.

Verification vs Trust: Why They Are Not the Same

A central claim of TMM is:

Verification confirms correctness under defined conditions. Trust, however, depends on:

• Provenance

• Context

• Institutional continuity

• Interpretation frameworks

The Five Pillars of Trusted Molecular Memory

TMM defines five interdependent pillars that together determine trust:

What Is the Molecular Chain of Custody (M-CoC)?

The Molecular Chain of Custody (M-CoC) tracks the lifecycle of a molecular artefact:

• Production

• Storage

• Measurement

• Transfer

• Regeneration

• Interpretation

Key Insight:
Each step can influence trust, even if the molecular data itself remains unchanged.

Scenario Analysis: When Valid Data Still Leads to Uncertainty

TMM introduces real-world-inspired scenarios to stress-test trust.

Scenario A: Regeneration and Material Continuity

• A voice message is encoded into DNA and stored.

• Later, a new DNA instance is regenerated from digital encoding.

• The decoded message is identical, but the molecule is different.

Conclusion:
Informational continuity ≠ material continuity.

Scenario B: Reading as Intervention

• DNA-stored archival material is periodically sequenced.

• Each read subtly alters the molecular material.

• Over time, the archive’s lineage shifts despite stable content.

Conclusion:
Measurement is not passive—it becomes part of custody.

Scenario C: Institutional Discontinuity

• A legal will is encoded in DNA and stored under formal supervision.

• Institutional changes occur (staff, systems, governance).

• The molecule remains intact and decodes correctly.

Conclusion:
Material stability ≠ institutional trust.

A New Failure Mode: Valid Data, Wrong Decision

Across all scenarios, TMM reveals a critical failure class:

This occurs when:

• Context is missing

• Provenance is unclear

• Institutional continuity is broken

• Interpretation frameworks are misaligned

Why AI Makes Molecular Trust More Urgent

AI systems are increasingly used to:

• Design biological sequences

• Interpret molecular outputs

• Support decision-making in bio-digital workflows

However:

• AI systems often treat data as ground truth

• Molecular data is context-dependent and historically situated

Key Risk:
AI can act on correct data but produce untrustworthy outcomes if it lacks context.

TMM Contribution:
Provides a framework for trust-aware AI systems that incorporate:

• Provenance tracking

• Contextual reasoning

• Institutional awareness

Why Biodesigners and Synthetic Biologists Should Care

For practitioners working with:

• DNA data storage

• Living materials

• Bio-digital interfaces

TMM introduces a critical design principle:

Practical Implications

• Design systems that track molecular lineage

• Treat measurement as an intervention, not observation

• Build institutional transparency into workflows

• Account for regeneration and transformation events

EEAT: Experience, Expertise, Authority, and Trustworthiness

This framework is grounded in interdisciplinary expertise across:

• Biodesign and synthetic biology

• Molecular data storage systems

• Human-computer interaction and design research

The work was developed in collaboration with:

• Dr. Larissa Pschetz (University of Edinburgh)

• Joe Revans (Fallow Earth)

• Prof. Thomas Heinis (Imperial College London)

Funded by the Advanced Research and Invention Agency (ARIA) under the Trust Everything Everywhere programme.

Key Takeaways

• Molecules store data, but trust emerges from context

• Verification alone is insufficient for decision-making

• Trust depends on five interconnected pillars

• Molecular systems introduce new failure modes

• AI systems must become context-aware to be trustworthy

FAQs About Trusted Molecular Memory (TMM)

What is Trusted Molecular Memory (TMM)?

TMM is a framework for evaluating trust in molecular systems by integrating material evidence, provenance, interpretation, accountability, and governance.

Why isn’t molecular verification enough?

Because verification confirms correctness, but not the context, history, or institutional validity of the data.

What is the Molecular Chain of Custody (M-CoC)?

A system for tracking how molecular artefacts are created, handled, transformed, and interpreted over time.

How does TMM relate to AI?

It helps AI systems move beyond pattern recognition to context-aware decision-making in biological workflows.

What is the biggest risk in molecular data systems?

The emergence of “valid data, wrong decision” scenarios due to missing or misunderstood context.

Conclusion: Trust Is a System Property, Not a Molecular Property

Trusted Molecular Memory reframes trust as something that emerges across systems—not something contained within a molecule.

In molecular computing and biodesign, trust is not stored. It is constructed.

Full Publication

Read the full paper:
https://doi.org/10.5281/zenodo.19002474

Somewhere in a laboratory freezer, a small tube holds a solution of synthetic DNA molecules. Encoded within their base sequences is a digital file, perhaps a medical record, a legal document, or a cultural heritage archive. To retrieve it, a researcher runs the sample through a sequencing machine, translates the molecular output back into binary, and reconstructs the original file. The process works. The data comes back intact. The system is considered to have succeeded.

But success at retrieval is not the same as being able to trust what was retrieved. That distinction is easy to overlook, and the DNA storage field has largely set it aside. This post argues that it should not.

What DNA Data Storage Actually Is

Every digital file, whether a photograph, a spreadsheet, or a genome sequence, is ultimately a string of ones and zeros. DNA, the molecule that encodes biological information in living cells, is also a string of information, written in a four-letter alphabet: adenine, thymine, cytosine, and guanine, abbreviated A, T, C, and G.

Binary data can be translated into sequences of DNA bases, synthesised as physical molecules in a laboratory, and stored until needed. To read the data back, the molecules are sequenced, their base order is read by a machine, and the result is translated back into the original digital file.

In simple terms: a digital file, such as a photograph, is converted into DNA, stored physically, and later converted back into the same photograph. If the reconstructed file looks identical to the original, the system is assumed to have worked correctly.

This assumption is what needs to be questioned.

DNA Is Not Just Digital. It's Physical.

Here is what makes DNA storage different from a hard drive or a cloud server in a way that matters beyond engineering detail. DNA is not an abstract medium. It is a molecule, and molecules exist in the physical world. They age. They react to their environment. They accumulate damage.

Every stage of the DNA storage pipeline leaves traces in the material. During synthesis, the process of writing data into DNA, certain sequences are produced in uneven quantities, introducing bias across the pool of molecules from the very beginning. During amplification, the process of making enough copies of the molecules to work with, that bias can compound, with some sequences replicated more reliably than others depending on temperature, chemistry, and protocol. During storage, the molecules degrade through chemical reactions: water breaks bonds, oxygen causes oxidative damage, and the longer the storage period and the less controlled the conditions, the more pronounced these effects become. Sequencing itself introduces characteristic error patterns. Contamination from prior samples, reagents, or the environment can persist across steps.

None of this is unusual or avoidable. These are intrinsic properties of working with DNA as a physical medium. Researchers who work with ancient DNA recovered from archaeological sites understand this intimately. They can read the damage patterns in a molecule and estimate how long ago the organism died, what conditions it was stored in, and whether the sample has been contaminated with more recently introduced material. The molecule's physical state carries a record of its own history.

In a DNA storage system, the same is true. The physical state of a sample at the point of retrieval reflects everything that has happened to it since it was first synthesised. That history is written into the molecule. The question is whether anyone is reading it.

What Current Systems Are Designed to Do

DNA storage systems have become technically sophisticated, and it is worth being precise about what they do well before identifying what they miss.

The central challenge in storing data as DNA is that the molecular channel is noisy. Synthesis makes mistakes. Sequencing makes mistakes. Some strands are lost during handling. To compensate, the field has developed robust error-correcting codes, mathematical schemes borrowed from telecommunications that add controlled redundancy to the data so that it can be reconstructed even when a significant fraction of molecules are damaged, missing, or misread. These codes work well. Studies have demonstrated error-free retrieval of DNA-encoded data after exposure to conditions designed to simulate decades of degradation. The codes absorb the physical damage and deliver the payload regardless.

Sequencing pipelines also include quality control steps that filter out unreliable reads before they enter reconstruction. Laboratory information management systems, known as LIMS, document the handling and provenance of samples at each stage of the workflow, creating a record of the process. Biobanking standards such as ISO 20387 formalise these requirements, mandating chain-of-custody documentation and traceability as part of quality management. Together, these layers constitute a mature technical stack, each component doing its job well.

But they share a structural limitation. Error correction ensures the data survives. Quality control ensures the sequencing reads are usable. LIMS and standards ensure the process is recorded. None of them ask a deeper question: does the physical state of the DNA match what the records claim about it?

A Failure Mode That Currently Goes Undetected

To make this concrete, consider a specific scenario.

A DNA sample is synthesised, encoded with data, and placed into long-term storage. The laboratory records note the synthesis date, the storage conditions, and the chain of custody. Years later, the sample is retrieved. The sequencing run proceeds without incident. The error-correcting code reconstructs the data without errors. The system reports success.

But suppose that at some point during those years, the sample was exposed to elevated humidity, perhaps during a facility move, a storage unit malfunction, or a lapse in protocol. Hydrolytic damage accumulated in the molecules. Some strands degraded. The population of surviving molecules became skewed toward the more stable sequences, while compromised ones dropped out. The error-correcting code, designed precisely for this kind of situation, compensated for the dropout and delivered the payload regardless.

The metadata still records stable storage conditions. The sequencing output still produces the correct data. No alarm was raised, because no part of the system was designed to compare the molecular evidence, the damage patterns, the skewed distributions, the dropout signatures, against the provenance record. The inconsistency is undetectable within the pipeline.

This is not a catastrophic failure. The data came back. But as a structural feature of how these systems are built, it creates a category of problem that is systematically invisible: cases where the physical reality of a sample and the recorded account of its history have diverged, and where retrieval success masks rather than resolves that divergence.

Other Fields Have Already Noticed This Problem

The idea that molecular signals constitute evidence about a sample's history, and that this evidence should be compared against recorded claims, is not new. It is standard practice in adjacent domains.

Researchers working with ancient DNA use damage patterns as a primary tool for authenticating samples. The characteristic chemical modifications that accumulate in DNA over centuries are specific enough to distinguish genuine ancient material from more recently introduced contamination, and to estimate how much contamination is present. The molecule's physical condition is treated as direct evidence about its provenance. In clinical and forensic genomics, similar logic applies: sequence-derived signatures are used to detect sample swaps, identify cross-contamination, and flag mismatches between what a sample's records say and what the molecular evidence implies.

These fields have developed the interpretive infrastructure to reason from physical evidence to trust assessments. DNA storage has not. The relevant signals are present in standard sequencing output, including error distributions, fragment length profiles, coverage patterns, and copy-number statistics, but they are currently discarded or collapsed into summary metrics that feed error correction, rather than being examined for what they reveal about the sample's condition and history.

What Is Missing

The gap is not in the data being produced. It is in what is done with it.

What current DNA storage systems lack is a way of treating molecular signals as evidence, comparing that evidence against recorded metadata, and forming a structured view of whether they are consistent. This is a different task from decoding. Decoding asks what information a molecule contains. The missing layer would ask whether the physical condition of that molecule matches the claims made about its history.

One way to think about the outcome of such a comparison is as a molecular trust state: a representation of how well the observed molecular evidence aligns with what is recorded about the sample.

This is not a binary judgement. Molecular evidence is inherently probabilistic. Different histories can produce similar patterns. Any meaningful assessment must express degrees of consistency and uncertainty rather than a simple pass or fail.

Trust, in this context, is not something a system either has or does not have. It is something that must be evaluated, with an understanding of what the evidence can and cannot resolve.

Why This Matters

For now, DNA storage is largely confined to research settings where experienced human oversight compensates for what the pipeline cannot detect. Researchers notice anomalies, flag unusual sequencing behaviour, and maintain institutional knowledge about particular samples and their histories. The system's blind spot is covered by people working alongside it.

That compensation does not scale. As DNA storage moves toward large institutional archives, cross-organisational data exchange, and records with legal, medical, or scientific weight, the pipeline becomes the authoritative system. Metadata from years or decades earlier is taken at face value. Retrieval success is treated as confirmation of integrity. The informal checks disappear, and what remains is a system that is very good at recovering data and structurally unable to evaluate the conditions under which that data should be trusted.

This shift is reinforced by a broader trend. Data pipelines are increasingly embedded in software systems that operate with minimal human oversight, including AI-driven workflows that rely on automated retrieval and processing. In these contexts, successful decoding is often treated as sufficient evidence of correctness. But automated systems do not question inconsistencies unless they are explicitly designed to do so. A DNA-based record can be retrieved perfectly and still carry physical signatures that contradict its recorded history, and no part of the system will raise that discrepancy.

Decoding success is not the same as trust. Metadata is not the same as truth. Molecular signals already carry information about the history and condition of a sample, and they are simply not being read for that purpose. Building the infrastructure to read them is not a matter of collecting new data. It is a matter of deciding that the question is worth asking.

The DNA storage field has made remarkable progress on retrieval. The next problem, verification, is different in kind, and the sooner it is recognised as a distinct engineering and design challenge, the better positioned the field will be to address it before the stakes make the oversight costly.

This is a research direction, not a finished answer. The questions it opens up, how to represent molecular condition in a structured way, how to reason probabilistically about sample history, how to build the comparison layers that a genuine molecular trust state would require, are genuinely hard and genuinely open. But they are the right questions to be asking now, before the infrastructure that ignores them becomes too entrenched to change.

If that intersection interests you, between biological materials, information integrity, and the design of trustworthy systems, this is a space worth watching. The work is early, the problems are real, and the field needs people willing to take the physical seriously alongside the digital.

Follow along as this research develops. New pieces go out through the Biodesign Academy newsletter. Subscribe here if you want them in your inbox.

Dear {{first_name | reader}},

Since last month’s announcement about the change in business structure, this is our first proper update on what has been happening.

One major project is now underway using national AI super-computing infrastructure. Another brings these questions into an international research forum in April. And a foundational research strand has recently concluded.

Across all of them sits the same concern:

As AI becomes embedded in biological design, how do we ensure what it generates is stable, accountable, and trustworthy?

Here is where things stand.

Stress-Testing AI-Designed Proteins for Biodesign

Our latest hands-on AI-biodesign project has been accepted through the UK’s AI Research Resource Gateway route and is now well underway.

I am serving as Principal Investigator on this project.

The work is supported through access to national AIRR supercomputing infrastructure, the Isambard-AI. The focus is AI-generated proteins relevant to biodesign and biofabrication.

Rather than optimising existing components, we are examining how newly generated biological interfaces behave under controlled variation. At this stage, the emphasis is not on performance improvement, but on structural stability.

As generative tools become embedded in design workflows, the question shifts from possibility to reliability. Not only what AI can generate, but how robust those outputs are when conditions change.

More details to follow.

CHI 2026 Panel

Biodesign × AI: Interactions in the Algorithmic Wet Lab

In April in Barcelona, I will be moderating a panel at CHI 2026 exploring what we call the algorithmic wet lab.

The panel brings together:

• Orkan Telhan

• Iohanna Nicenboim

• Margherita Pevere

• Carolina Ramirez-Figueroa

AI systems are no longer only analysing biological data. They are shaping experimental direction, material design, and optimisation workflows.

That raises structural questions:

Who designs when AI participates in laboratory decisions? Where does responsibility sit when outcomes emerge from human–algorithm–organism systems? What does accountability look like when interpretation is distributed?

As moderator, my role is to surface tensions between technical capability, design responsibility, and institutional governance.

More information about the session will follow shortly.

Rethinking Trust at the Molecular Scale

The ARIA-supported Trusted Molecular Memory project has now concluded.

The work formed part of Advanced Research and Invention Agency’s Trust Everything, Everywhere discovery programme, examining how trust infrastructures must evolve as computation intersects with physical and biological systems.

The project was led by Larissa Pschetz at the University of Edinburgh. Through Biodesign Academy, I contributed on the analytical side, focusing on how verification must be reframed when AI-driven automation meets material processes.

Rather than proposing a new technical mechanism, the work concentrated on clarifying how trust is structured in hybrid bio-digital systems.

Further details will be shared once the publication is publicly available.

Independent Technical Diligence

Across these strands of work, a consistent theme emerges.

As AI enters biological design, claims accelerate. Stability, feasibility, and governance do not automatically follow.

Alongside research, I am conducting independent technical diligence for EU and UK investors and institutions evaluating AI-driven biology ventures.

The focus is straightforward: separating biological feasibility from narrative-driven AI claims.

If you are assessing AI–biology companies or institutional AI integration in this space, feel free to get in touch.

AI in biology is moving quickly. What matters now is not only capability, but clarity about how these systems hold up under scrutiny. The projects above are different expressions of the same effort: building reliability into AI-driven biodesign.

Until next time,

Raphael

Founder, Biodesign Academy

How AI Is Reshaping Biodesign: Stability, Accountability, and Trust in Algorithmic Biology

As AI becomes embedded in biological design, the central question is no longer what it can generate, but whether those outputs are stable, accountable, and trustworthy under real-world conditions.

This update outlines three interconnected strands of work:

• Stress-testing AI-designed proteins using national supercomputing infrastructure

• Examining responsibility in the “algorithmic wet lab” at CHI 2026

• Rethinking trust in hybrid bio-digital systems through ARIA-supported research

• Conducting independent technical diligence on AI-driven biology ventures

Together, they address a single structural issue: how to build reliability into AI-driven biodesign.

Stress-Testing AI-Designed Proteins Using National AI Infrastructure

What happens when AI generates biological components from scratch?

A new hands-on AI–biodesign research project is now underway via the UK’s AI Research Resource (AIRR) Gateway route, with access to national supercomputing infrastructure: Isambard-AI.

I am serving as Principal Investigator on this project.

Research Focus

• AI-generated proteins relevant to biodesign and biofabrication

• Structural stability under controlled variation

• Behaviour of newly generated biological interfaces

• Reliability under changing environmental conditions

Unlike optimisation-focused workflows that refine existing proteins, this project investigates first-principles generation — examining how de novo AI-designed biological components behave when conditions shift.

Why Stability Matters

As generative systems move into laboratory workflows:

• Possibility is no longer the main concern

• Reliability becomes the core issue

• Robustness under perturbation determines feasibility

In biotechnology and biofabrication contexts, small instabilities can cascade into material failure or experimental collapse. AI outputs must therefore withstand scrutiny beyond computational metrics.

This work contributes empirical data to a growing field that often advances faster in claims than in verification.

Biodesign × AI at CHI 2026: The Algorithmic Wet Lab

In April 2026, I will moderate a panel at CHI 2026 (Barcelona) titled:

Biodesign × AI: Interactions in the Algorithmic Wet Lab

The session explores how AI systems are not merely analysing biological data but actively shaping:

• Experimental direction

• Material design

• Laboratory optimisation workflows

Panel Participants

• Orkan Telhan

• Iohanna Nicenboim

• Margherita Pevere

• Carolina Ramirez-Figueroa

These researchers and practitioners work across design, HCI, and bio-digital systems.

Core Structural Questions

When AI participates in laboratory processes:

1. Who is designing?

2. Where does responsibility sit?

3. How is accountability structured when outcomes emerge from human–algorithm–organism systems?

4. What governance mechanisms apply to distributed interpretation?

My role as moderator is to surface tensions between:

• Technical capability

• Design responsibility

• Institutional governance

As AI moves from analytical tool to experimental co-agent, responsibility frameworks must evolve accordingly.

Rethinking Trust at the Molecular Scale

How should trust function in hybrid bio-digital systems?

The Trusted Molecular Memory project, supported by the Advanced Research and Invention Agency (ARIA) under its Trust Everything, Everywhere discovery programme, has now concluded.

The project was led by Larissa Pschetz at the University of Edinburgh.

Through Biodesign Academy, I contributed analytical work focused on:

• Reframing verification in AI-driven automation

• Examining how trust infrastructures adapt when computation intersects with physical and biological systems

• Clarifying structural trust mechanisms in hybrid environments

Key Insight

Rather than proposing a new technical control layer, the project examined:

• How trust is constructed

• Where verification breaks down

• Why existing computational trust models fail at molecular and material scales

As biological systems become programmable, verification must extend beyond digital audit trails into material processes.

Publication details will follow upon release.

Independent Technical Diligence in AI-Driven Biology

Why diligence matters in AI–biology ventures

Across research and advisory work, a consistent pattern emerges:

• Claims accelerate faster than feasibility

• Stability is assumed rather than tested

• Governance lags behind integration

Alongside academic research, I conduct independent technical diligence for EU and UK investors and institutions assessing AI-driven biology companies.

Diligence Focus Areas

• Biological feasibility vs. AI narrative inflation

• Structural stability of generated biological systems

• Material verification pathways

• Governance and regulatory risk

• Integration into institutional infrastructure

This work draws on extensive experience across biotech, design academia, and industry collaboration — bridging computational methods and biological material systems.

The goal is simple: separate capability from credibility.

Building Reliability into AI-Driven Biodesign

AI in biology is advancing rapidly. The technical frontier is expanding.

What now matters most is:

• Structural robustness

• Verification under variation

• Clear responsibility models

• Trust infrastructures aligned with material reality

The projects outlined above — national supercomputing research, international research dialogue, and trust-focused inquiry — represent different expressions of the same effort:

Ensuring AI-driven biodesign systems hold up under scrutiny.

Key Takeaways

• AI-generated proteins must be tested for structural stability, not just computational plausibility

• Algorithmic participation in laboratory design raises accountability challenges

• Trust in bio-digital systems requires new verification frameworks

• Independent technical diligence is critical as AI–biology ventures scale

• Capability alone does not ensure feasibility or governance readiness

FAQs About AI in Biodesign

What is AI-driven biodesign?

AI-driven biodesign refers to the use of generative and analytical AI systems to design biological materials, proteins, interfaces, and laboratory workflows.

Why is stability more important than optimisation?

Optimisation improves performance within known systems. Stability determines whether newly generated biological systems remain functional under real-world variation.

What is the “algorithmic wet lab”?

The algorithmic wet lab describes laboratory environments where AI systems actively influence experimental direction, design choices, and material outcomes.

Why does trust need to be rethought at the molecular scale?

Traditional computational trust models rely on digital verification. Biological systems introduce material uncertainty, making verification more complex.

What does independent technical diligence involve in AI–biology?

It involves assessing biological feasibility, structural robustness, governance frameworks, and alignment between technical claims and material realities.

Biodesign is often framed as a more ethical, more sustainable way of working with the living. But what if the very systems that make biodesign possible rely on vast forms of labour that remain largely unseen, unnamed, and unaccounted for? In this issue, we are sharing a conversation with Yuning Chen, a biodesigner and PhD researcher in Design Informatics at the University of Edinburgh, whose work asks exactly that question.

Through her Labour Provenance framework, Chen traces the hidden ecologies of organisms that sustain everyday laboratory practice, from microbes and enzymes to the infrastructures that keep them alive. In the interview below, she reflects on her path into biodesign, why labour became a critical lens for her work, and what it might mean to take more-than-human contributions seriously in biodesign practice.

Yuning Chen working in the lab (image by Y.Chen)

Can you briefly introduce your background and how you came into the intersection of HCI, design informatics, and biodesign?

My background is in environmental science and later design engineering. I started my undergraduate studies with a somewhat naive ambition to heal the world, only to be met with the disenchantment that it is often more common to develop complex technological fixes for pollution than to challenge cultures of consumerism, wastefulness, and environmental indifference. That realisation led me to take design courses as a creative way to ask different questions, not only how to solve problems, but why we create them through the ways we live, consume, and relate to the more-than-human world.

During my design masters, I was naturally drawn to working with living organisms. I was fascinated by how their agential capabilities evade control and certainty. I saw biodesign practices as a way to mediate a fairer ground for negotiation between humans and other forms of life. However, my brief experience in the biotech industry exposed a more pragmatic side. I began to notice a familiar pattern, a passion for innovation and constant newness that promises a better, more sustainable future, yet often reinscribes the same logic of extraction that caused our current crisis.

With this tension, I returned to academia. In Design Informatics and HCI, I encountered a community of researchers asking questions that resonated with my unease toward biotechnological narratives. What happens if we stop putting humans at the centre of technological design? How might technologies, infrastructures, and practices be reshaped if we take more-than-human perspectives seriously? From my perspective, the expanded field of HCI is not only about interfaces between humans and computers, but also about the cultural and political interfaces between humans, technologies, and the more-than-human world. That is where I found grounding for my critical biodesign practice.

Plant Reality Set: early biodesign work of Chen that explores human attunement to plant (image by Y.Chen)

What drew you to questions of more-than-human ecologies and labour in the first place?

Ecologies come quite naturally to me. Perhaps due to my undergraduate training, I have always tended to think of organisms in networks and assemblages. What is striking is how easily this ecological lens becomes backgrounded in laboratory and biodesign practices. In these contexts, organisms are often presented in purified monocultures, or in engineering biology, as collections of genetic Lego blocks. This creates a false sense of individuality and detachment.

Only when one looks behind the curtain, asking where enzymes, sera, or cell lines come from, where products go, who sustains the lab, and who provides nutrients for growth, does it become clear that both organisms and labs are embedded in wider ecologies. For me, this is precisely why it is important to bring an ecological lens into practices dominated by purity and decontextualisation.

Labour was a lens I first encountered through Despret and Porcher’s writing about cows. They described how interpreting organisms’ failures as resistance can reveal their active participation and labour in agricultural production. I was drawn to how labour offers an alternative way to question relationships with organisms and demand justice. Framing organisms as labouring subjects is not necessarily about intentionality, but about recognising agency in political and economic terms, extending ethical reflection into questions of structural oppression.

What motivated you to look at biodesign through the lens of labour provenance?

I came across the term organism agnosticism, which describes how engineering biology seeks to make living systems engineerable in ways that are agnostic to their origins. This made me curious about the abstraction of living entities from ecological and evolutionary contexts into standardised genetic blocks.

Initially, I focused on how feral organisms were transformed into homogenous chassis or blocks. Through conversations with biology colleagues, I realised that many everyday lab reagents and tools, such as enzymes and polymerases, also have organismal origins. As I investigated further, I kept discovering new clusters of living sources behind tools and materials in my own experiments.

I decided to trace the provenance of lab reagents used in a project involving common gene-editing techniques to understand the hidden ecologies sustaining daily synthetic biology lab operations. At the time, I was inspired by Wadiwel’s work on animals and capital, which analyses production systems through the lens of labour. That framework became crucial for making sense of the structural roles organisms occupy within biotechnological production, and that is how the labour provenance method emerged.

A labour provenance map (non-exhaustive) based on a biodesign experiment (image by Y.Chen)

Could you walk us through the hybridisation experiment, why yeast and human cells, and what you hoped to unsettle by pairing them?

In this experiment, we use hybridisation in an expanded sense of pairing and connecting two organisms through surface adhesion, rather than fusing them into a single organism. The hybridisation of yeast and human cells, two entities with very different moral status, was intended to unsettle the boundaries of people’s moral imagination.

Yeast is a long-domesticated microbial worker and is rarely present in ethics discussions. Human cells, in contrast, are morally charged and evoke concerns about exploitation, consent, or cannibalism. Pairing them entangled these moral registers and positioned human cells as witnesses to the microbial labour that underpins daily life and scientific discovery.

Human cells were used as a tactical entry point for what Acampora calls corporal compassion, a form of compassion grounded in shared bodily vulnerability. Through this pairing, we hoped to mediate compassion toward the labour and vulnerability of other living entities in the lab.

A microscope image of yeast and HEK cell co-culture experiment (image by Y.Chen)

The paper identifies five types of more-than-human labourers. Which of these surprised you the most?

My level of surprise was almost inverse to the order of the five labour types, with evaluation labour being the most surprising and primary labour the least. This was less about the importance of their contribution and more about their visibility.

I began by looking into cell line histories and commonly used genes, which are primary and specific skills labourers. It was discovering the organisms behind gene-editing tools, sustenance, and even gene ladders that prompted me to ask more systematically where these abstractly named components come from.

I was surprised to find that not only tools, but also sustenance systems and evaluation mechanisms, can all be derived from living sources. This revealed how deeply laboratory operations appropriate existing ecological relations, and how quickly visibility and recognition decline from the centre to the periphery of biotech production.

The five more-than-human labourers (image by Y.Chen)

How did participants in your workshop respond to mapping organisms as workers with job titles?

Participants were generally surprised by the sheer number of organisms involved in their work, starting with main experimental organisms and gradually extending to more assistive roles. It was rewarding to see everyday reagents reframed not simply as tools, but as outcomes of living ecologies.

When assigning job titles, language played a significant role. Descriptive, playful, or metaphorical names conferred different degrees of agency and recognition. The final question about how to treat more-than-human employees better revealed a pluralistic understanding of ethical relevance. This reinforced my view that ethical sensibilities toward less emotionally relatable organisms can be rehabilitated through recognising labour and interdependence.

Sample of workshop notes (image by Y.Chen)

Where do you see the biggest blind spots in biodesign’s sustainability claims?

One of the biggest blind spots is equating biodesign directly with sustainability. This risks treating organisms as regenerative sources of functionality without considering the labour and physical pressures they are subjected to, or the wider labour ecologies that support them.

Not viewing sustainability as contingent on more-than-human labour ecologies is, for me, a significant limitation within biodesign discourse.

How do you think this framework can influence biodesign practice in labs?

I see Labour Provenance as a call to action across individual practices, tools, protocols, and institutional procurement. For individual researchers, it broadens sustainability considerations by revealing the multitude of life implicated in everyday experiments, and introduces a new ethical dimension around organismal labour and subsumption.

For research communities, it could foster cultures of appreciation and memorialisation that extend beyond primary organisms. For institutions, it could inform procurement policies by demanding transparency around the organismal sources of lab materials, which could in turn create pressure for suppliers to account for embedded labour.

Image by Y.Chen

How do you balance academic critique with practical design application?

As a practice-based researcher, I see critique and practice as co-constitutive. Without working in the lab, I would not have arrived at this perspective. Without theory, I would not have been able to contextualise and articulate it.

Practice provides the site where tensions emerge, while theory offers the lens to make them legible and consequential. My next step is to explore how these reflections can translate into relational and methodological shifts within biodesign practice.

Chen and her art exhibition showcasing part of the supply chain ecologies behind biotechnology (image by Y.Chen)

Looking ahead, what kinds of collaborations would you most like to see emerging from this work?

I would like to see collaborations that bring together biologists, designers, STS scholars, technologists, regulators, and procurement officers. The most important contribution of Labour Provenance is broadening sustainability and ethics discourse in biodesign, which requires collaboration across these different strata.

I am interested in designers working with procurement managers to visualise labour provenance, biologists collaborating with ritual designers to create memorials, and ecologists interpreting the artificial ecosystems produced by lab practices and supply chains. By revealing the entanglement of biodesign, supply chains, and multispecies labour, Labour Provenance could offer common ground for recognising shared responsibilities and acting toward more accountable forms of design with the living.

A fluorescent bread crumb from the bread made of human-yeast hybrid cells, which is part of Chen’s collaborative art project with synthetic biologists (image by Y.Chen)

This Is Where Biodesign Gets Uncomfortable

Before you close this email, it may be worth pausing on a simple question that runs quietly through this entire conversation. Who is really working in the lab? Yuning Chen’s work does not ask us to abandon biodesign or to resolve its ethical tensions neatly. Instead, it invites us to notice the layered, more-than-human labour ecologies that make biological design possible in the first place. By making these relations visible, Labour Provenance opens space for more careful, accountable, and situated ways of working with the living. As biodesign continues to grow, that shift in attention may be just as important as any new material or technique.

Thank you, Yuning for sharing your work, below are her social links if you want to check out her work further and to get in touch:

Dear {{first_name | reader}},

Biodesign has long been limited by what biology already evolved. Materials like mycelium, algae, and bacterial cellulose were mostly explored through craft, observation, and slow experimentation, even as more rational bioengineering methods emerged in parallel.

Over the past five or so years, that landscape has shifted: not because experimentalists suddenly became better, but because modern AI systems now expose structure and pattern at scales that were previously inaccessible.

Below are three grounded ways AI is expanding what designers consider possible, anchored in real research rather than hype.

1. Making hidden structure legible

Biology is hard to design with because much of its structure is invisible at the scale of intuition. AlphaFold2, introduced in 2021, changed that by delivering highly accurate protein structure predictions for a vast range of sequences, and by making those predictions accessible through the AlphaFold Protein Structure Database. This did not replace experiments, but it did give scientists and designers a way to inspect predicted folds, motifs, and interfaces before stepping into the lab.​

AlphaFold DB now hosts well over 200 million predicted protein structures derived from UniProt-scale sequence sets, taking structural coverage from hundreds of thousands of experimental models to hundreds of millions of inspectable 3D forms.

By 2023–2024, many labs working on enzymes and binders routinely used AlphaFold2 predictions and confidence metrics (such as pLDDT and PAE maps) to decide which designs or constructs were worth testing experimentally. For designers, this is analogous to the way CAD made mechanical components legible: biological “shape” and potential interaction surfaces became something you can browse, filter, and reason about at unprecedented scale.​

A parallel at the material scale comes from mycelium composites. Studies have shown that mechanical and physical properties (such as compressive stiffness, density, thermal conductivity, and water absorption) depend systematically on substrate type, fibre condition, and processing.

These are precisely the kinds of structure–property relationships that AI and statistical models can help explore, not by replacing experimentation, but by narrowing which combinations are worth making. What was once opaque is becoming legible, and legibility expands imagination.​

2. Exploring “what if?” beyond natural precedent

AI now enables systematic exploration of biological variations that would have been slow or impractical to probe experimentally at scale. In protein design, this shift is especially visible in de novo generative methods.

ProteinMPNN provides deep-learning-based sequence design on specified protein backbones, while diffusion-based models such as RFdiffusion generate entirely new backbone geometries and scaffolds. Together with newer frameworks like Chroma, these tools allow researchers to propose new shapes, topologies, and functional surfaces that do not occur in natural proteins, and to evaluate them in silico before committing to experimental work.​

At the same time, updated versions of AlphaFold (such as AlphaFold3) are extending prediction beyond single proteins to complexes with DNA, RNA, and small molecules, enabling in silico exploration of many candidate interactions that would be difficult to screen solely by hand or traditional docking workflows.

For biodesigners, this means questions like “What if this protein folded differently?”, “What if this interface bound a different partner?”, or “What if this scaffold reinforced a living material?” can now be explored far more systematically and at much greater scale than was realistic five years ago. The important nuance is that de novo design existed before, but current AI systems make these explorations broader, faster, and more accessible.​

3. Bringing speculation closer to feasibility

Historically, speculative biodesign concepts often sat far from biological constraints; many ideas could not be checked without extensive, slow experimentation. AI now helps close that gap, not by guaranteeing outcomes, but by acting as an early plausibility filter.

In mycelium composites, for example, properties such as stiffness, density, and water absorption have been shown to correlate with measurable parameters like substrate type, fibre size, fungal species, and incubation or processing conditions.

Across materials science more broadly, machine-learning models (including random forests, neural networks, and convolutional architectures) are already used to learn structure–property relationships and to pre-filter candidate formulations or process windows before fabrication. For bio-based composites like mycelium, this suggests a near-term path where AI helps select promising regions of parameter space, while the actual materials are still validated through growth and testing.​

Bacterial cellulose (BC) offers a similar story. Experimental and review papers show that BC thickness, porosity, and mechanical performance respond in systematic ways to culture conditions such as oxygen availability, vessel geometry, medium composition, and strain engineering. While designer-facing AI tools specific to BC are still emerging, general modeling and ML approaches can, in principle, help explore how boundary conditions and geometries might influence growth and properties, guiding which “what if” samples to prioritize.​

Most strikingly, AI-designed binders and mini-proteins generated by workflows based on models like RFdiffusion have been expressed in the lab and shown to bind their targets with high affinity, including recent examples where AI-designed binders modulate biological pathways or improve genome editing efficiency. These pipelines do not eliminate experimental work, but they demonstrate a clear path from computational imagination to physical, functional molecules.​

For designers, this changes the nature of speculation. The goal is not perfect prediction, but triage: identifying which ideas are impossible, which are merely interesting, and which are plausibly worth a growth cycle or a set of assays. AI provides early signals that used to require full experimental loops to obtain.​

Why this shifts biodesign’s horizon

When new structures become visible, new ideas become plausible. When new variations can be explored at scale, new design spaces open up. When feasibility can be tested computationally before committing to lab work, imagination can expand without losing touch with constraints.

AI does not replace biodesign practice; it reconfigures it. The field is moving from a practice largely bounded by natural precedent and hand-tuned craft toward one animated by computational imagination, in which designers navigate vast spaces of possible forms, behaviours, and interactions, then bring a carefully chosen subset into the lab to see what biology will actually do.​

Until next time,

Raphael

Biodesign Academy

Stay up-to-date with AI

The Rundown is the most trusted AI newsletter in the world, with 1,000,000+ readers and exclusive interviews with AI leaders like Mark Zuckerberg, Demis Hassibis, Mustafa Suleyman, and more.

Their expert research team spends all day learning what’s new in AI and talking with industry experts, then distills the most important developments into one free email every morning.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.

Dear {{first_name | reader}},

Enrolment for the AI × Biodesign course is now finally open.

The first cohort starts 9 February 2026, and we are sharing the full curriculum, tools, and guest details on our dedicated page.

The course is designed to give designers a practical, grounded way into molecular reasoning, living materials, and AI-driven biology.

If you’re working with biodesign, materials, speculative practice, or emerging technologies, this will give you the literacy and clarity the field increasingly demands.

We’ll cover molecular behaviour, AI structure interpretation, living-material logic, feasibility, and design workflows across six live sessions.

Recordings are included, and the cohort is limited to 15 seats.

Early-access price: €299 until Tuesday, 25 November 2025 (23:59 CET).
Standard price: €349 afterwards.

If you have questions about fit or content, just reply, I’ll get back to you personally.

Raphael
Founder, Biodesign Academy

Dear {{first_name | reader}},

AI is moving into biology fast, and it's hard to separate the signal from the noise.

Every month there's a new model, a new claim, a new "breakthrough." Some of it matters. A lot of it is just good marketing.

But something real is shifting.

Tools that used to require institutional access: structure prediction, sequence design, material modelling, are becoming more available.

You no longer need a PhD, a lab, or the right academic connections to work with these ideas.

That opens up new possibilities. It also raises new questions.

Designers need to understand how these tools actually work, where they fail, and how to use them without getting caught up in the hype. That's not optional anymore, it's practical.

To help with that, I'm opening early access to a new six-week cohort next week:

AI × Biodesign: A Practical Course in Molecular Workflows for Creatives (February 2026)

This is a focused introduction to molecular design and AI-supported biological reasoning, built for creatives who want clarity, not confusion.

Why now

Biotech has always been mostly-gated: paywalled journals, specialist labs, closed datasets, academic gatekeeping. That's changing.

More preprints. More open models. More accessible tools. More ways in.

The barrier to entry is dropping, if you know where to look and how to interpret what you find.

At the same time, the hype is creating a different problem: inflated expectations, overinterpreted results, design decisions made without understanding the biology underneath.

This programme sits in the middle: accessible and grounded, but scientifically honest about what these tools can and can't do.

What we'll cover

Over six weeks, we'll build a foundation in:

• Reading protein structures and molecular behaviour

• Using AI tools like AlphaFold in a biodesign workflow

• Understanding how enzymes influence material properties (texture, colour, density, response)

• Applying molecular logic to mycelium, bacterial cellulose, and algae

• Extracting useful information from scientific papers using GPT workflows

• Translating molecular behaviour into materials and forms

• Navigating the ethical and speculative dimensions of AI-driven biology

The goal isn't to make you a scientist. It's to give you fluency and good judgement in a field that's moving quickly.

Tools you'll use

Three core resources throughout the programme:

• Designer AlphaFold Notebook

• Living Material Proxy Library

• Biodesign Co-Pilots (Notion + AI workflows)

Built for long-term use, not just exercises.

Two expert collaborators

To keep things technically sound while staying designer-friendly, two experts will join us:

• A computational biologist specialising in protein modelling

• An expert materials researcher working with living organisms

Their input is practical and tied directly to the tools we'll use.

Who this is for

Creative practitioners, design students, designers, interaction/industrial/architecture practitioners, creative researchers in living materials, early-stage founders exploring biological systems.

Beginner-friendly, but serious.

Early access opens Wednesday

The cohort is capped at 15 to keep sessions small and interactive.

On Wednesday morning, you'll get:

• Full syllabus

• Tool previews

• Collaborator details

• Early-access registration link

Programme starts February 2026.

More soon,
Raphael
Biodesign Academy

Choose the Right AI Tools

With thousands of AI tools available, how do you know which ones are worth your money? Subscribe to Mindstream and get our expert guide comparing 40+ popular AI tools. Discover which free options rival paid versions and when upgrading is essential. Stop overspending on tools you don't need and find the perfect AI stack for your workflow.

Subscribe to Get Your Free Comparison

AI × Biodesign: A Practical Course in Molecular Workflows for Creatives

A six-week live cohort, starting February 2026

AI is moving into biology at high speed. Each month brings a new model, a new claim, a new “breakthrough.” Some advances genuinely shift what’s possible; others are polished marketing. The important signal: capabilities that once required specialist labs, institutional access, or deep academic training are becoming broadly available. Designers and creatives can now work directly with molecular ideas — if they understand how to use these tools responsibly.

Why This Matters Now

Biotech has long been gated by paywalls, specialist labs, and academic hierarchies. That landscape is loosening.

• Open preprints and transparent methods

• Freely accessible or low-cost AI models for prediction and design

• Tools that rely on intuition and reasoning rather than credentials

• More ways to experiment without institutional affiliation

As access expands, hype accelerates. Misinterpreting AI outputs leads to unrealistic expectations and design decisions detached from real biology. Creative practitioners now need a grounded understanding of both potential and limitation.

What This Course Provides

This six-week programme offers a focused introduction to molecular design and AI-supported biological reasoning. It prioritizes clarity, practical workflows, and sound judgement.

Core Topics

• Reading protein structures and molecular behaviour

• Using AI tools like AlphaFold in design-led workflows

• Understanding how enzymes influence texture, colour, density, and responsiveness

• Applying molecular logic to mycelium, bacterial cellulose, and algae

• Extracting essential insights from scientific papers using GPT workflows

• Translating molecular behaviours into materials and forms

• Navigating ethical and speculative dimensions of AI-driven biology

The aim is competency and fluency, not scientific credentialing.

Tools You’ll Use

Three core resources support the programme:

• Designer AlphaFold Notebook

• Living Material Proxy Library

• Biodesign Co-Pilots (Notion + AI workflows)

These tools are designed for long-term use beyond the cohort.

Expert Collaborators

Two domain specialists help keep the programme scientifically rigorous and materially relevant:

• A computational biologist specialising in protein modelling

• A materials researcher working directly with living organisms

Their contributions connect directly to the workflows taught.

Who It’s For

Ideal participants include:

• Design students

• Speculative designers

• Interaction, industrial, and architectural practitioners

• Creative researchers in living materials

• Early-stage founders exploring biological systems

Beginner-friendly, but serious in depth and application.

Early Access Opens Wednesday

The cohort is capped at 15 participants to maintain an intimate, interactive environment.

On Wednesday morning, participants will receive:

• Full syllabus

• Tool previews

• Collaborator details

• Early-access registration link

The programme begins February 2026.

Summary

AI is expanding who can participate in molecular and biological design. With new access comes the need for new literacy: understanding how these tools work, how to interpret their limits, and how to apply them with precision. This programme provides that foundation for creatives navigating a rapidly evolving field.

FAQs

What can AI actually do in biodesign today?

It enables rapid structure prediction, sequence exploration, and modelling of molecular behaviour — capabilities once restricted to specialist labs. These support material exploration, speculative design, and early-stage concepts.

Do I need a scientific background?

No. The course is structured for designers. Concepts are taught from first principles and grounded in hands-on tools.

How is this different from a standard biotech or computational biology course?

It is design-oriented and workflow-driven. The focus is reasoning, interpretation, and applied creative practice.

What will I be able to do afterward?

You’ll be able to read protein structures, interpret molecular behaviour, use AI tools responsibly, evaluate biological claims, and integrate molecular reasoning into your projects.

Why limit the cohort to 15?

Small groups enable deeper discussion, technical guidance, and personalised feedback.

More soon,
Raphael
Biodesign Academy

Dear {{first_name | reader}},

Last week at the Next Nature Conference in Eindhoven, I shared my talk Responsible AI × Biodesign — exploring how artificial intelligence is changing how we design with life.

Many of you asked to see the full slides and notes, so here they are. This is actually the full slide deck, which is extended version used for the the talk itself, so it has additional slides and notes:

Download the full deck + presenter notes (PDF)

The presentation covers:

• How AI models like AlphaFold and ESMFold are reshaping molecular design

• Why responsibility, empathy, and uncertainty are new design materials

• What it means to design within thresholds — between algorithms and organisms

For Foundational Tier members, you now get extra resources:

• Editable Powerpoint & Google Slides version you can adapt for your own teaching or projects

• Biodesign Academy ImageBank — 170 curated visuals and diagrams used in the talk and other past materials. Feel free to use them, and give credit to Biodesign Academy when appropriate.

Not a member yet? You can upgrade to Foundational Tier here: Join Now

These resources were created to help you bring responsible, AI-driven biodesign into your own classroom, lab, or studio — without needing a lab coat or supercomputer.

Thank you for the energy and curiosity following my talk.
Let’s keep building a responsible, imaginative future: one fold, one molecule, one idea at a time.

Raphael Kim
Biodesign Academy

Choose the Right AI Tools

With thousands of AI tools available, how do you know which ones are worth your money? Subscribe to Mindstream and get our expert guide comparing 40+ popular AI tools. Discover which free options rival paid versions and when upgrading is essential. Stop overspending on tools you don't need and find the perfect AI stack for your workflow.

Subscribe to Get Your Free Comparison

Dear {{first_name | reader}},

Next Friday, I’ll be speaking at the Next Nature Conference 2025 – Bio Design, part of Dutch Design Week at the iconic Evoluon in Eindhoven.

The conference gathers designers, scientists, and artists exploring how AI, living systems, and sustainable materials are shaping a regenerative future.

Preparing for my talk — “What Does Responsible AI Look Like in Biodesign?” — made me realise how often we still think of artificial intelligence as something synthetic, detached, or even opposed to life.

But perhaps AI is not the opposite of nature.

Perhaps it is one of its continuations.

Reframing the “Artificial”

Biology and computation are both learning systems. One evolves through enzymes and proteins, the other through data and code.

When we teach an algorithm to model a protein, it performs a kind of digital evolution: iterating, adapting, optimising, failing, trying again. It’s not so different from how nature explores the possible.

If we accept that, then the question shifts. It’s no longer should AI belong in biodesign?
It becomes how do we work with it responsibly?

Responsibility as a Material

In biodesign, responsibility is a material — as real as agar, cellulose, or mycelium.

Every model we train and every biological system we engineer carries assumptions about value, purpose, and ownership.

Working with AI and living systems side by side means confronting those assumptions early — before they calcify into practice.

To design responsibly, we have to think ecologically about our algorithms: who they serve, what they consume, and what kinds of futures they help us imagine.

Toward a More-Than-Human Collaboration

The most interesting question in this space is not what AI can do for biology, but what biology can teach us about AI — adaptability, empathy, circular logic.

Designing responsibly with AI might not be about controlling life, but learning from it.
And perhaps, the real intelligence to strive for is the one that cares.

I’ll share my slides, references, and a reflection piece after the conference next week, with Foundational Tier members getting access much earlier than that.

Until next time,

Raphael

Biodesign Academy

What Does Responsible AI Look Like in Biodesign?
Exploring the Continuum Between Nature and Intelligence

Introduction: When AI Meets Life Itself

Artificial intelligence is often described as something synthetic — detached, mechanical, and opposed to the organic world. Yet, when we look closer, AI and biology share a fundamental principle: both are systems that learn, adapt, and evolve.

As I prepare to speak at the Next Nature Conference 2025 – Bio Design during Dutch Design Week at the Evoluon in Eindhoven, one question continues to guide my thinking:

Reframing the “Artificial”

Biology and computation mirror one another as learning systems:

• Biology evolves through enzymes, proteins, and mutations.

• Computation evolves through data, models, and iterations.

When an algorithm learns to model a protein, it performs a kind of digital evolution — iterating, adapting, and optimizing just as life does.

This reframing shifts the central question:

Responsibility as a Design Material

In biodesign, responsibility itself is a material — as tangible as agar, cellulose, or mycelium.

Every trained model or engineered organism embeds assumptions about value, purpose, and ownership. Working with AI and living systems side by side requires surfacing those assumptions before they calcify into default practice.

Responsible AI design means thinking ecologically about algorithms:

• Who do they serve?

• What do they consume?

• What futures do they help us imagine?

Ethical design in this context isn’t a constraint; it’s a form of material intelligence.

Toward a More-Than-Human Collaboration

The deeper question is not what AI can do for biology, but what biology can teach AI:

• Adaptability – evolution’s way of handling complexity.

• Empathy – understanding relational systems, not isolated agents.

• Circular logic – designing processes that regenerate rather than deplete.

Designing responsibly with AI may not mean controlling life but learning from it. The most advanced intelligence may not be the one that dominates — but the one that cares.

Summary: A Regenerative Vision for AI and Biodesign

Key Takeaways

• AI and biology share learning-based logics — both evolve through feedback and adaptation.

• Responsibility should be treated as a core material in biodesign, shaping every stage of AI development.

• Ethical AI design in biodesign means creating systems that regenerate, not just automate.

• True intelligence may emerge through empathy, care, and ecological understanding.

FAQs About Responsible AI in Biodesign

Q1. What does “responsible AI” mean in the context of biodesign?
It refers to designing AI systems that consider ecological, ethical, and social impacts — ensuring that computational processes align with the principles of living systems.

Q2. How can designers demonstrate responsibility as a material?
By embedding ethical reflection and transparency in the design workflow — from dataset selection to lifecycle evaluation of biotechnologies.

Q3. Why is reframing “artificial” important?
Because the term implies separation from life, while in reality, both AI and biology are part of a continuum of adaptive systems.

Q4. What is “more-than-human” collaboration?
It’s the practice of designing with awareness of nonhuman systems — understanding that intelligence extends beyond the human to the biological and ecological.

Conclusion: Designing the Intelligence That Cares

As we design the next generation of biological and computational systems, our challenge is not just to make them powerful, but responsible.

The future of AI in biodesign may depend less on control — and more on care, adaptability, and ecological intelligence.

Author: Raphael
Speaking at the Next Nature Conference 2025 – Bio Design, Dutch Design Week, Eindhoven

Tags: Responsible AI, Biodesign, Bioinformatics, Ethical Design, Sustainability, Regenerative Futures

The Safety Information Crisis in Biodesign

Dear {{first_name | reader}},

A biodesign student spends six hours searching for Spirulina safety protocols. A startup delays production for three weeks because they can't verify containment requirements for mycelium.

A studio contaminates their workspace because no one knew Bacillus subtilis needed autoclaved disposal.

The information exists, buried in 200-page PDFs written for compliance officers, not designers building the future of materials.

The Biodesign Safety Co-Pilot solves this in 30 seconds.

After watching too many talented designers abandon projects because they couldn't access basic safety information, I built the tool I wished had existed. This is for everyone who believed biology was "too regulated" to explore freely.

→ Generate Your First Safety Brief

What You Get

Enter any organism or biomaterial. Receive a structured brief with:

• Biosafety Level (BSL) and Risk Group classification

• Required PPE and containment measures

• Waste disposal and environmental protocols

• Plain-language context for your workspace

Copy it into your SOP. Share it with your team. Actually understand what you're doing.

Photo by Levart_Photographer on Unsplash

Why Not Just Ask ChatGPT?

Generic AI gives inconsistent answers and won't distinguish E. coli K-12 (safe) from O157:H7 (pathogenic).

The Safety Co-Pilot uses specialized prompting refined over 50+ iterations to deliver:

• Standardized safety categories every time

• Risk-appropriate language for designers, technicians, or companies

• Disposal specifics often missing from generic AI

• Built-in verification prompts for institutional review

It's the difference between "information" and "guidance you can use."

Real Example: Koji Mold

Before: Designer googles Aspergillus oryzae. Finds sake forums and a 2004 paper. Unclear if ventilation needed. Studio has no autoclave…is that a problem?

After 30 seconds: BSL-1, GRAS status. Work in ventilated area, no biosafety cabinet needed. Wear N95 for dry spores. Dispose with 10% bleach (30 min). If scaling beyond 5L, consult institutional biosafety.

Now they know: Current studio works with ventilation. N95 masks sufficient. Bleach replaces autoclaving. Exactly when to escalate.

This clarity should have existed from day one.

How It Works

Runs entirely in your browser. No backend, no tracking, no data storage.

1. Paste your OpenAI API key (never stored)

2. Enter organism or material

3. Select your role (Designer/Technician/Company)

4. Get your brief in ~30 seconds

Foundational Tier members receive full source code to audit, modify, and deploy internally.

Photo by Coasteering on Unsplash

Trust & Limitations

This tool provides educational guidance, not institutional authorization.

Always verify with your institution's EHS office for GMO work, consult MSDS for chemicals, and escalate when scaling beyond research quantities.

Beta limitations: Best for well-documented organisms. Does not replace formal risk assessments. Focuses on biological hazards, not chemical safety.

We built transparency into the architecture because trust requires verifiability.

Try These Three

1. Spirulina platensis — GRAS organisms still need specific handling

2. Mycelium (Pleurotus ostreatus) — Spore exposure risks for "safe" fungi

3. Bacterial cellulose — Disposal requirements for fermentation

Each search teaches you something generic AI misses.

→ Start Now

In Closing…

Safety is foundational. This is another tool in a suite designed to make specialized biological knowledge accessible without sacrificing rigor. Together, they form an intelligence infrastructure for creative practitioners working with living systems.

Until next week,

Raphael Kim
Founder, Biodesign Academy

FAQs About the Biodesign Safety Co-Pilot

What is the Biodesign Safety Co-Pilot?

The Biodesign Safety Co-Pilot is a browser-based AI tool that generates structured, lab-ready biosafety briefs for any organism or biomaterial in about 30 seconds. It tells you what PPE to use, how to handle waste, and whether your workspace meets safety requirements—before you start working.

How does the Safety Co-Pilot improve biosafety in design practice?

It transforms scattered institutional safety data into clear, actionable guidance. Instead of searching hundreds of pages of PDFs, users instantly receive risk classifications, containment steps, and disposal protocols specific to their context (designer, technician, or company).

Why can’t I just use ChatGPT or a generic AI model for biosafety?

General AI tools are inconsistent—they often blur safe and pathogenic species, omit PPE guidance, or misclassify biosafety levels.
The Safety Co-Pilot uses a custom prompt architecture refined through over 50 iterations and aligned with WHO/CDC and EHS standards, producing verified and structured safety briefs.

Who should use the Safety Co-Pilot?

It’s designed for anyone working with living materials—designers, educators, students, lab technicians, startups, and studios. Each role receives the appropriate level of safety depth, from beginner-friendly analogies to full SOP-format protocols.

How does it generate results without storing my data?

The Co-Pilot runs entirely in your browser. It uses your OpenAI API key locally, never uploads or stores data, and keeps all operations on the client side. There’s no backend or tracking, ensuring full privacy.

What kind of information do the safety briefs include?

Each brief covers four structured areas:

1. Regulatory Classification — Biosafety Level (BSL) and Risk Group (RG).

2. Practical Handling Protocol — PPE, containment, and decontamination steps.

3. Waste & Environmental Guidance — Disposal and release procedures.

4. Plain-Language Context — Why each step matters for your project.

Is the tool open source or auditable?

Yes. Foundational Tier members can access the complete open-source codebase to audit prompts, verify methods, or adapt the tool to institutional standards. Transparency and verifiability are core design principles.

How accurate is the information provided?

The tool’s guidance is derived from validated biosafety standards (CDC, WHO, EHS). It’s most accurate for well-documented species and research-scale contexts but should always be verified locally for GMO or scale-up work.

What are its known limitations?

• Works best for common organisms; rare species may produce general guidance.

• Not a substitute for institutional risk assessments or grant documentation.

• Focuses on biological, not chemical, safety.

• Training data current through January 2025—users should confirm any reclassifications.

How does this tool fit into Biodesign Academy’s larger mission?

The Safety Co-Pilot is the first part of Biodesign Academy’s intelligence infrastructure—AI companions that make biosafety, materials performance, and ecological design accessible to all. Future tools include the Materials, Protein, and Regeneration Co-Pilots, all built on the same open and educational framework.

What makes this approach trustworthy?

Transparency, verifiability, and user control. Every prompt can be inspected, every result explained, and every action runs locally. The Co-Pilot models a new kind of responsible AI—one built on biosafety literacy rather than black-box automation.

How can I start using it?

Visit Generate Your First Safety Brief to enter your organism or biomaterial name.
No sign-up required, no data tracking—just 30 seconds to get the safety clarity your project needs.

Dear {{first_name | reader}},

I wanted to share a project that’s been quietly taking shape behind the scenes.

Meet the Biodesign Co-Pilot: Safety, an AI-assisted tool that generates instant biosafety briefs for organisms and biomaterials.

Whether you’re working with yeast, bacterial cellulose, or engineered enzymes, the aim is simple: before you design, check the safety.

The Co-Pilot summarises biosafety level, risk group, handling guidance, and potential environmental considerations, in seconds rather than hours.

It’s part of a broader goal at Biodesign Academy: making responsible biodesign fast, clear, and integrated into creative practice.

Safety shouldn’t slow innovation down; it should inform it. By putting biosafety context next to design intent, we make more intelligent and defensible decisions.

Here’s how it works:

• You enter any organism or material — for example E. coli K-12 MG1655 or chitosan film.

• The Co-Pilot returns a concise “Safety Brief” with containment level, handling notes, and references to EU/WHO/NIH guidelines.

• It finishes with a short reflection prompt: “If this were scaled beyond the lab, how would you design for containment?”

The Safety Co-Pilot will launch next week for Foundational Tier Members, with downloadable PDF briefs and integration into the Biodesign Academy archive.

If you’d like early access, you can upgrade here.

I’d also love to know which organisms or materials you’d like supported first, just reply to this email and let me know.

Until next week,
Raphael
Biodesign Academy

Structured Summary:

What Is the Biodesign Safety Co-Pilot and How Does It Make Biosafety Fast and Intelligent?

Overview: A Tool for Instant Biosafety Insight

The Biodesign Safety Co-Pilot is an AI-assisted platform that generates instant biosafety briefs for organisms and biomaterials.
Developed by Biodesign Academy, it helps designers, scientists, and students check biosafety risks before they begin a project — turning compliance into a creative, informed design step rather than a bottleneck.

Why the Biodesign Safety Co-Pilot Matters

Modern biodesign often moves faster than regulatory review or safety assessment. The Co-Pilot solves this gap by integrating biosafety knowledge directly into the creative workflow.

• Goal: Make responsible biodesign fast, clear, and actionable.

• Approach: Deliver trusted biosafety information — instantly, in context.

• Impact: Designers can make safer, more defensible choices from the start.

How the Biodesign Safety Co-Pilot Works

The system analyzes biological inputs and returns structured biosafety guidance in seconds.

Step-by-step process:

1. Enter an organism or material — e.g., E. coli K-12 MG1655, yeast, chitosan film, or bacterial cellulose.

2. Receive a concise Safety Brief including:

   • Biosafety level and risk group

   • Handling and containment guidance

   • Environmental and ethical considerations

   • Source references (EU, WHO, NIH guidelines)

3. Reflect and design responsibly: Each brief includes a prompt such as
   “If this were scaled beyond the lab, how would you design for containment?”

Key Benefits for Designers and Researchers

• Speed: From hours of manual research to seconds per query

• Accuracy: Draws on verified biosafety frameworks and regulatory databases

• Clarity: Summarizes risk data in plain, actionable language

• Integration: Exports results to the Biodesign Vault for project documentation

• Reflection: Encourages a mindset of proactive biosafety awareness

Access and Launch Details

The Biodesign Safety Co-Pilot launches next week for Precision Members of Biodesign Academy.
Subscribers will gain:

• Access to downloadable PDF briefs

• Integration with the Biodesign Vault

• Early input on new organism and material support

To request early access, visit the Biodesign Academy membership page.

Summary

FAQs About the Biodesign Safety Co-Pilot

What kind of data does the Co-Pilot use?

It references international biosafety standards including EU, WHO, and NIH frameworks for biological classification and containment.

Who can use it?

Open to Biodesign Academy members working in synthetic biology, biomaterials, or design research. No advanced coding or regulatory expertise required.

How accurate are the Safety Briefs?

Each summary is algorithmically generated but grounded in verified biosafety documentation and human-reviewed databases.

How is this different from a traditional biosafety manual?

Instead of static documents, the Co-Pilot offers dynamic, query-based guidance that adapts to the designer’s material or organism of interest.

Can users suggest new organisms or materials?

Yes — members are invited to submit requests for inclusion in upcoming database updates.

Author’s Note

Written by Raphael of Biodesign Academy, drawing on firsthand experience in biodesign education, biotech practice, and responsible design systems. The article reflects the Academy’s ongoing effort to integrate safety, creativity, and ethics into emerging bio-innovation workflows.

Dear {{first_name | reader}},

Over the past weeks you’ve been getting in-depth biodesign tools from us in the form of copilots, but not last week and not this week… I’ve been in the middle of relocating from the Netherlands to Sweden with my family.

Still, I didn’t want to leave you without something useful. I’d also promised a while back to give you a simple guide to publishing on Zenodo, so here it is: a step‑by‑step walkthrough of how to share your biodesign work on Zenodo, with examples tailored to materials like mycelium, bacterial cellulose, and algae.

Why Zenodo matters for biodesign

Biodesign sits between science, design, and art. Traditional journals and conferences are often geared toward finished research papers or polished presentations.

Zenodo, by contrast, lets you share work-in-progress, teaching packs, protocols, and raw datasets in a formal, citable way. That makes it easier for others to build on your work while it’s still evolving.

• Journals: usually accept full papers after peer review, often long delays, and not ideal for raw data or workshop materials.

• Conferences: great for visibility and community, but your talk or poster rarely gets a DOI and can be hard to find afterwards.

• Zenodo: immediate, free, and DOI-backed. You can upload anything from tensile strength data of mycelium sheets to algae cultivation protocols, and others can cite it just like a paper.

Zenodo doesn’t replace journals or conferences. Instead, it complements them by giving your background work, materials, and teaching resources a proper record.

The DOI system (made simple)

When you upload something to Zenodo, it gives you a DOI — that’s a Digital Object Identifier, basically a permanent link that makes sure your work can always be found and cited.

There are actually two kinds. The first is the concept DOI, which always points to the latest version of your project, no matter how many times you update it.

The second is the version DOI, which is unique to each specific upload. If you shared data on mycelium growth today and then update it next month with new results, each upload gets its own version DOI.

When you or anyone else cites the work, it’s best to use the version DOI so readers know exactly which set of files you’re referring to.

Getting started

The first step is simply to visit the Zenodo website: https://zenodo.org. From there you can create an account and begin uploading your work.

Five steps for biodesign projects

1. Create an account and connect ORCID
   ORCID is a free researcher ID that makes sure your outputs follow you, even if you move between labs, studios, or schools.

2. Choose a license

   • Code (e.g., for simulation scripts) → MIT or Apache-2.0.

   • Data/docs (e.g., lab protocols, workshop packs) → CC BY 4.0.

3. Organise your files
   Structure matters. For example:

   • /data – growth measurements of bacterial cellulose films

   • /code – scripts for analysing mycelium density

   • /docs – workshop notes on algae cultivation

   • README.md, CITATION.cff, and LICENSE

4. Fill in the details

   • Title: descriptive, e.g., “Dataset: Tensile Strength of Mycelium Composites (v1.0.0)”.

   • Description: 150–400 words explaining what it is, how it was created, and how others can reuse it.

   • Keywords: think like a searcher — biodesign, mycelium, bacterial cellulose, algae, biomaterials.

   • Related links: link your GitHub repo, project website, or paper.

   • Funding: include grant or sponsor info if relevant.

5. Version properly
   Biodesign projects evolve. Label releases clearly (v1.0.0, v1.1.0, etc.). Each new version gets its own DOI.

Examples in practice

• Mycelium: upload microscopy images of growth patterns with a README on sample prep.

• Bacterial cellulose: share tensile strength data from dried sheets, with code for plotting stress–strain curves.

• Algae: publish pigment concentration data and workshop protocols for classroom experiments.

• Workshops: create a Zenodo record for teaching packs with slides, handouts, and Arduino sketches — all with a DOI so students can cite them.

Optional: connect GitHub

If you’re prototyping with code (e.g., simulation of growth, generative design scripts), link your GitHub to Zenodo. Every new GitHub release will mint a DOI automatically.

Things to avoid

• Uploading without ORCIDs → weak attribution.

• No license → unclear if others can use your work.

• Vague description → harder for future reuse.

• Dumping files without README → nobody knows what’s inside.

• Citing only the concept DOI → always cite the version you used.

Templates you can copy

README header

# <Project Title>

Short summary: what it is, who it’s for, why it matters.

## Contents

- /data – raw and processed CSVs

- /code – analysis scripts

- /docs – methods + figures

## How to cite

Use the version DOI: <10.5281/zenodo.xxxxxx>

CITATION.cff (minimal)

cff-version: 1.2.0

message: "If you use this project, please cite it."

title: <Project Title>

version: <1.0.0>

doi: <10.5281/zenodo.xxxxxx>

authors:

- family-names: <Surname>

given-names: <Name>

orcid: "https://orcid.org/<ORCID>"

Bottom line

Biodesign thrives on sharing across labs, studios, and classrooms. Journals and conferences remain essential for finished, peer‑reviewed work, but Zenodo is different: it makes everything around your project: datasets, protocols, workshop packs, even images, citable and permanent.

Think of it as the missing layer that ties your practice together and keeps it discoverable.

And so that’s it for this week, short but hopefully practical. If you’ve never uploaded anything to Zenodo before, consider trying it with one of your smaller biodesign outputs, whether it’s a dataset on algae growth or a workshop handout on bacterial cellulose.

Once you’ve done it once, you’ll see how quick it is, and your work will instantly become easier to share, cite, and build on. More in‑depth copilots will return soon, but for now I hope this guide gives you something concrete you can use right away.

Until next week,

Raphael, Biodesign Academy

^{Structured Summary:}

How to Publish Biodesign Projects on Zenodo: A Step-by-Step Guide

Why Zenodo Matters for Biodesign

Biodesign projects often sit between science, design, and art. Traditional publication routes have limits:

• Journals: accept only full peer-reviewed papers, with long delays, unsuitable for raw data or workshop materials.

• Conferences: offer visibility but talks and posters rarely get a DOI, making them hard to cite later.

• Zenodo: free, immediate, DOI-backed. Allows sharing datasets, protocols, images, and teaching packs in a formal, citable way.

👉 Zenodo complements journals and conferences by giving your background work, evolving materials, and teaching resources a permanent, discoverable record.

Understanding DOIs (Digital Object Identifiers)

When you upload to Zenodo, you receive a DOI that ensures your work remains findable and citable:

• Concept DOI → always points to the latest version of a project.

• Version DOI → unique to each upload; preferred for citations so readers know exactly which version you used.

Example: upload tensile strength data of mycelium composites today, update next month with new results → each dataset has its own DOI. Always cite the version DOI.

How to Get Started with Zenodo

1. Create an account and connect ORCID
   ORCID ensures your contributions follow you across labs, studios, and institutions.

2. Choose the right license

   • Code → MIT or Apache-2.0

   • Data and teaching materials → CC BY 4.0

3. Organize your files
   Example structure:

   /data – growth measurements of bacterial cellulose films
   /code – scripts for analysing mycelium density
   /docs – workshop notes on algae cultivation
   README.md, CITATION.cff, LICENSE
   

4. Fill in project details

   • Title: descriptive and versioned (e.g., Dataset: Tensile Strength of Mycelium Composites v1.0.0)

   • Description: 150–400 words explaining purpose, method, reuse potential

   • Keywords: biodesign, mycelium, bacterial cellulose, algae, biomaterials

   • Related links: GitHub, project website, publications

   • Funding: acknowledge grants or sponsors

5. Version properly
   Use semantic versioning (v1.0.0, v1.1.0, etc.). Each release generates a new DOI.

Examples of Biodesign Projects on Zenodo

• Mycelium: microscopy images of growth patterns with README on sample prep.

• Bacterial cellulose: tensile strength datasets plus plotting scripts.

• Algae: pigment concentration data and classroom workshop protocols.

• Workshops: full teaching packs with slides, handouts, and Arduino sketches.

Optional: connect GitHub → each release automatically mints a new DOI.

Common Mistakes to Avoid

• Uploading without ORCID → weak attribution.

• No license → unclear reuse rights.

• Vague descriptions → harder for future researchers.

• Dumping files without README → unusable datasets.

• Citing only the concept DOI → always cite the version DOI.

Ready-to-Use Templates

README Header

# <Project Title>

Short summary: what it is, who it’s for, why it matters.

## Contents
- /data – raw and processed CSVs
- /code – analysis scripts
- /docs – methods + figures

## How to cite
Use the version DOI: <10.5281/zenodo.xxxxxx>

Minimal CITATION.cff

cff-version: 1.2.0
message: "If you use this project, please cite it."
title: <Project Title>
version: <1.0.0>
doi: <10.5281/zenodo.xxxxxx>
authors:
  - family-names: <Surname>
    given-names: <Name>
    orcid: "https://orcid.org/<ORCID>"

Bottom Line

Zenodo strengthens biodesign by making evolving work — from datasets to teaching packs — citable, permanent, and discoverable. Journals and conferences remain vital for polished, peer-reviewed research, but Zenodo adds the missing layer for sharing experimental data, protocols, and educational resources.

If you’ve never uploaded before, start with something small, like an algae growth dataset or a bacterial cellulose workshop handout. The process is fast, and your work immediately becomes easier to share, cite, and build upon.

FAQs About Zenodo for Biodesign

Q: Is Zenodo free to use?
Yes, Zenodo is completely free, supported by CERN and the European Commission.

Q: Can I update my dataset later?
Yes. Each version gets its own DOI, while the concept DOI points to the latest version.

Q: Do I need an ORCID to publish on Zenodo?
It’s not mandatory but strongly recommended for attribution and long-term visibility.

Q: What file types can I upload?
Anything: datasets, images, code, teaching packs, videos, PDFs.

Q: Does Zenodo replace journals and conferences?
No, it complements them by making supporting material accessible and citable.

Dear {{first_name | reader}},

There’s a certain shimmer to showcase halls.. beautiful booths, brilliant lighting, biomaterials poised like sculptures. But after the lights fade, what remains is the maker’s question: How do I actually begin? This week, we offer a different kind of stage: one without applause, but full of traction.

The Biodesign Lab-to-Market Co-Pilot is not a spectacle (yet), but more of an AI-driven, practical tool. Built for those ready to move from idea to action, it transforms vision into structure, with just enough friction to test what’s real.

Introducing the Biodesign Co-Pilot: Lab-to-Market

Turn a biomaterials idea into a decision-grade plan in minutes.
This isn’t a showcase but a working tool you can use today.

What it does

Paste your concept (e.g., “mycelium acoustic panels for offices”) and select a region (EU/US/UK). The Co-Pilot returns a source-backed plan with:

• Regulatory path — jurisdiction-aware frameworks and test methods (e.g., EN 13501-1, ISO 10993, FDA 510(k) overview) with citations.

• Go-to-market wedge — a narrow beachhead, buyer archetypes, where to find them, and a first-contact email draft.

• 12-week pilot plan — milestones with exit criteria.

• Risk ledger — severity/likelihood scoring, mitigations, and kill criteria.

• Unit economics (rough) — COGS range, target price, and a pilot budget estimate.

• Citations index — links so you can verify anything quickly.

Why it matters

Fairs inspire. You need decisions. The Co-Pilot converts research and intent into a credible route to first customers, with enough structure to brief a supervisor, partner, or investor.

How to use it

1. Open the Co-Pilot (with your own OpenAI API key at the ready).

2. Enter your idea, choose GEO, add optional cost inputs.

3. Generate the plan, then export PDF/CSV/JSON for your notes or deck.

Roadmap

• Saved reports and shareable links

• Expanded standards directory by domain

• Light team features (comments, versioning)

For Foundational members

You’ll soon get access to the source code and our embedded regulatory/test knowledge base to adapt the Co-Pilot for your own projects. Watch this space.

If you use it to run a pilot, reply with a line on your results. We’ll feature the strongest examples.

In a field enamored with promise, we choose to walk beside the process. The Co-Pilot isn’t designed to dazzle, it’s meant to serve. Quietly, steadily, it supports the work that outlives the exhibit: the prototypes built, the pilots run, the risks taken and learned from.

Until next week,
Raphael, Biodesign Academy

What Is the Biodesign Co-Pilot and How Does It Help Turn Lab Ideas Into Market Plans?

Overview

The Biodesign Co-Pilot: Lab to Market is an AI-driven practical decision-support tool that transforms biomaterials concepts into structured, source-backed plans in minutes. Instead of being a showcase, it provides a working framework for founders, researchers, and innovators navigating the lab-to-market journey.

Key Features of the Biodesign Co-Pilot

• Regulatory Path
  Jurisdiction-aware guidance (EU/US/UK) with frameworks, test methods, and citations.
  Examples: EN 13501-1, ISO 10993, FDA 510(k) overview.

• Go-to-Market Wedge
  Identifies buyer archetypes, beachhead markets, outreach channels, and even generates a first-contact draft email.

• 12-Week Pilot Plan
  Provides milestones, exit criteria, and structured deliverables for an initial pilot.

• Risk Ledger
  Scores risks by severity and likelihood, suggests mitigations, and defines kill criteria.

• Unit Economics
  Rough COGS estimates, target pricing, and a pilot budget range.

• Citations Index
  All results are source-backed, with links for quick verification.

Why It Matters

• Fairs inspire ideas, but innovators need decision-ready plans.

• The Co-Pilot converts research and intent into credible routes to first customers.

• Outputs are structured enough to brief a supervisor, partner, or investor.

• It supports practical execution beyond inspiration: prototypes, pilots, and risk management.

How to Use the Co-Pilot

1. Open the Co-Pilot (requires an OpenAI API key).

2. Enter your biomaterials concept.

3. Choose a target region (EU/US/UK).

4. Add optional cost inputs.

5. Generate and export results as PDF, CSV, or JSON.

Note: Outputs are advisory; regulatory and testing standards must be validated with an accredited lab or notified body.

Roadmap

• Saved reports and shareable links

• Expanded standards directory by domain

• Team features (comments, versioning)

• Foundational members gain early access to source code and embedded regulatory/test knowledge base

Try It

👉 Biodesign Co-Pilot: Lab to Market

If you use it to run a pilot, share your results. The best examples will be featured by Biodesign Academy.

FAQs About the Biodesign Co-Pilot: Lab to Market

What is the Biodesign Co-Pilot: Lab to Market?
An AI-driven decision-support tool that transforms biomaterials concepts into structured, source-backed market entry plans.

Who should use it?
Founders, researchers, and innovators in biomaterials, biotech, and biodesign.

Does it provide regulatory guidance?
Yes, tailored by jurisdiction (EU/US/UK) with citations to relevant frameworks and standards.

Can it estimate costs?
It provides rough unit economics, including COGS ranges, target prices, and pilot budgets.

Is it production-ready or advisory?
Outputs are advisory and must be confirmed with accredited labs or notified bodies.

What’s next for the tool?
Upcoming features include saved reports, shareable links, an expanded standards directory, and team collaboration tools.

Dear {{first_name | reader}},

Ok, so maybe the headline was a bit too much, but today marks an interesting turning point for Biodesign Academy.

Biodesign Academy started by shaping the conversation around biodesign. Now we’re shaping the tools. The launch of our first Co-Pilot marks the beginning of a new phase: AI-powered software built for educators, students, and practitioners.

Not a CustomGPT. Not a prompt library. Actual software, designed to support biodesign workflows with speed, rigor, and clarity.

Release 1: Biodesign Co-Pilot: Ideation

The first Co-Pilot tackles the hardest part of any project: turning vague curiosity into a structured, testable idea.

With just three inputs: 1) a biological system, 2) a property, and 3) an application domain, it generates a project brief complete with:

• A clear concept hook

• Mechanisms & precedents to explore

• Safety & ethics notes, classroom-ready

• A practical bill of materials

• Timeboxed methods you can follow

• Evaluation criteria for outcomes

• Variants for low-resource classrooms

• Further reading prompts

This is a guided, structured assistant that turns possibility into action.

Co-Pilot: Ideation in Action

Biodesign Co-Pilot: Ideation — a browser-based assistant that generates structured project briefs from simple inputs. Users enter their own OpenAI API key, select biology, property, and domain, and the Co-Pilot outputs a workshop-ready concept.

(Biological System dropdown):
Choosing a biological system to work with. Here, the user highlights “Fungi (mycelium)” as the starting point for ideation.

(Application Domain dropdown):
Picking the design context where the project will be applied. “Bioelectronics” is selected as the focus area.

(Audience dropdown):
Defining the target audience. Here, “educators (course planning)” is chosen, tailoring the output for teaching contexts.

Users can hit generate button, or the Surprise Me button to get randomized output

Here’s the Output

To show you what the Co-Pilot can do, we ran it with the following inputs:

• Biological System: Algae

• Target Property: Biodegradability

• Application Domain: Fashion

• Audience: Beginner design students

• Constraints: BSL-1 only, <$100 budget, 2-week build

The result? A structured, workshop-ready project brief that can be used in classrooms, prototyping sessions, or early-stage research. See result summary PDF here.

Mini Guide: Trying the Co-Pilot Yourself

You can try Co-Pilot: Ideation today [Link here] by bringing your own OpenAI API key.

I know this is a bit of a pain, but once you have the key it is straightforward to use.

Here’s how:

1. What’s an API key?
   It’s like a personal password that lets software (like this Co-Pilot) talk to OpenAI’s models. Each key is linked to your OpenAI account and usage.

2. How do I get one?

• Go to https://platform.openai.com/account/api-keys.

• Log in or sign up.

• Click Create new secret key.

• Copy the key (it starts with sk-...).

3. Paste it into the Co-Pilot.
   Your key is stored locally in your browser — it isn’t sent to us. You control it, and you pay only for your usage (a few dollars’ credit lasts a long time with lightweight models).

4. Try it out.
   Pick your biology, property, and domain → hit Generate → get a structured project brief.

⚠️ Important: Don’t share your API key publicly. Treat it like a password.

No API Key? No Problem.

If you don’t want to set up an API key, you don’t have to. From next week, Biodesign Academy Foundational Tier members get access to a hosted version of the Co-Pilot with no setup required.

👉 [Join Biodesign Academy Foundational Tier Membership]

What’s Next

Co-Pilot: Ideation is just the beginning. We’re already developing:

• Co-Pilot: Publication — simplify making your work citable.

• Co-Pilot: Safety — instant safety briefs for organisms and materials.

• Co-Pilot: Materials — map biology to practical applications.

• Co-Pilot: Teacher — lesson plans, rubrics, and assessments auto-generated from your projects.

Over time, the Co-Pilot Series will cover the entire biodesign workflow.

This is a big step forward for Biodesign Academy. We’ll continue to provide toolkits, prompt libraries, and CustomGPTs — but Co-Pilots are something else entirely. They are software products, not just resources.

Anyone can spin up a CustomGPT. Very few are building purpose-built AI assistants for biodesign education. This is the first of its kind — and it’s only the start.

Until next time,

Raphael, Biodesign Academy

❓ FAQ: Biodesign Co-Pilot: Ideation

What is Biodesign Co-Pilot: Ideation?
Biodesign Co-Pilot: Ideation is an AI-powered web app that generates structured biodesign project briefs. It turns user inputs (biology, property, application domain, audience) into a ready-to-use concept with methods, safety notes, and references.

How is Biodesign Co-Pilot different from a CustomGPT?
Unlike a CustomGPT or prompt library, the Co-Pilot is a dedicated software product with structured outputs designed for education and prototyping. It’s built specifically for biodesign workflows.

Who can use Biodesign Co-Pilot?

• Beginner and advanced design students

• Educators creating lesson plans and workshops

• Bioengineering students and HCI researchers

• Startups prototyping biodesign projects

What output does Biodesign Co-Pilot create?
The Co-Pilot generates:

• Concept and rationale

• Safety and ethics guidance

• Bill of materials

• Step-by-step methods

• Evaluation criteria

• Low-cost classroom variants

• Further reading references

Do I need an OpenAI API key to use Co-Pilot?
No — there are two ways to use it:

1. Bring Your Own API Key (connect your own OpenAI account).

2. Biodesign Academy Membership (hosted access, no setup required).

How do I get an OpenAI API key?
Go to platform.openai.com/account/api-keys, log in, and create a new key. Copy it immediately and paste it into the Co-Pilot. Keys are reusable, and usage credits determine costs.

What are the next Co-Pilots coming after Ideation?

• Co-Pilot: Publication – create metadata and submission packs

• Co-Pilot: Safety – generate classroom biosafety briefs

• Co-Pilot: Materials – connect biological inputs to applications

• Co-Pilot: Teacher – auto-generate lesson plans and rubrics

How do I access Biodesign Co-Pilot: Ideation?
Sign up for Biodesign Academy Membership for immediate access to Co-Pilot: Ideation and all future Co-Pilot tools.

Why is this launch important for Biodesign Academy?
This is Biodesign Academy’s first AI software product. It marks a move beyond newsletters and toolkits into building practical, purpose-built AI assistants for biodesign education and practice.

Dear {{first_name | reader}},

If you’ve ever worked on a biodesign project, you’ll know the feeling:
You want feedback on your concept, maybe it’s an early prototype, maybe a speculative object, maybe a material process still rough around the edges. But the moment you hit send or show it to someone outside your immediate circle, doubts creep in:

• Am I giving away the core of my idea?

• Will this get misinterpreted, copied, or judged too early?

• How do I invite critique without opening up the lab notebook entirely?

This hesitation is a very biodesign-specific tension. Unlike traditional design, our work often sits on top of biological detail: strains, sequences, plasmids, fabrication protocols. Things that feel both essential and sensitive.

That’s why we built AnonBio: a simple, client-side tool that helps you anonymize your ideas before sharing them.

What is AnonBio?

Think of it as a filter for your design process notes:

• Identity scrub: Removes personal names, labs, emails, URLs, and grant numbers that could tie a concept too tightly to you or your collaborators.

• Technical softening: Swaps out exact biological details — like strain IDs, plasmid names, or concentrations, with placeholders ([organism], [plasmid], [concentration]). Enough for the design intent to come through, but not the raw recipe.

• Risk radar: Flags if your text still contains too much sensitive data.

• Quick preview: See what’s been swapped out at a glance, or copy the “safe” plain text version.

• Email composer: Draft a message around your anonymized text and send it for feedback, all without leaving your browser.

All of this happens in your browser window. No servers, no logins, no data leaving your machine.

Why this matters for biodesign

Biodesign thrives in conversation. Our projects are rarely linear, they cross between lab, studio, and speculation. We build ideas by bouncing them between contexts, between people, between disciplines.

But the fear of leakage often interrupts that flow. A student might hold back in a critique session. A researcher might avoid sharing a speculative experiment with designers. A startup founder might skip a valuable design conversation in case the IP seeps out.

AnonBio is a response to that blockage.
It doesn’t solve every issue of trust or ownership, but it creates a lightweight ritual: scrub first, then share. By anonymizing, you create a safer space for feedback, without losing the essence of what you’re trying to explore.

Try it

Free vs. Foundational Tier

• Free readers → You get the tool itself, ready to use right away.

• Foundational Tier members → You’ll also get an extended kit:

  • The full source code (so you can adapt it to your own studio/lab workflow).

  • An annotated breakdown of how the anonymization patterns work.

  • Reflections on how this fits into biodesign processes — from studio crits to lab handovers to speculative publishing.

Looking ahead

AnonBio is just the first experiment. Imagine:

• Studio crits where early biodesign ideas can circulate safely.

• A future peer review mode where speculative projects can be assessed without prejudice.

• Workshop formats where sensitive material can be anonymized before collective mapping.

We’d love to hear how you would use this. Does it open up conversations you’d otherwise avoid? What contexts in your practice could benefit from it?

Reply and let us know — this is a living experiment, just like biodesign itself.

Best wishes,

Raphael, Biodesign Academy

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

How AnonBio Helps Biodesigners Share Ideas Safely

What Problem Does AnonBio Solve?

Biodesign projects often face a dilemma:

• How can designers and researchers share early concepts without exposing sensitive lab details?

• How can feedback be invited without risking misinterpretation, copying, or premature judgment?

• How can conversations flow when intellectual property or biological data might leak?

Unlike traditional design, biodesign projects often involve biological details such as strains, sequences, plasmids, and fabrication protocols. These elements feel essential yet sensitive. This creates hesitation in sharing ideas openly.

AnonBio was created as a solution: a lightweight tool that anonymizes biodesign notes before sharing them.

What AnonBio Does

AnonBio works as a client-side anonymization filter for design process notes. Its main functions include:

• Identity scrub → Removes names, labs, emails, URLs, and grant numbers to protect contributors.

• Technical softening → Replaces biological details (strain IDs, plasmid names, concentrations) with placeholders like [organism] or [plasmid].

• Risk radar → Alerts users if sensitive data remains.

• Quick preview → Shows what has been swapped and generates a plain-text safe version.

• Email composer → Lets users draft and send anonymized feedback requests directly.

✅ Importantly, all processing happens locally in your browser. No servers, no logins, no data leaving your machine.

Why AnonBio Matters for Biodesign

Biodesign thrives on conversation across labs, studios, and disciplines. However, fear of data leakage often interrupts collaboration.

• Students may hesitate to share during critiques.

• Researchers may avoid speculative discussions with designers.

• Startup founders may withhold design insights to protect IP.

AnonBio introduces a ritual of anonymization: scrub first, then share. This protects sensitive details while preserving the design intent, creating safer feedback spaces.

How to Try AnonBio

You can test the tool online:

1. Paste your concept note.

2. Toggle “conceal technical details.”

3. Preview the anonymized version and copy safe text for sharing.

Free vs. Foundational Tier

• Free access → Use the live anonymization tool immediately.

• Foundational Tier membership → Unlocks extended resources:

  • Full source code for adapting workflows.

  • Annotated patterns of anonymization.

  • Reflections on how anonymization supports biodesign processes, from critiques to publications.

Looking Ahead: Future Uses of AnonBio

AnonBio is the first step in experimenting with safer biodesign collaboration. Potential applications include:

• Studio critiques where early ideas can circulate without risk.

• Peer review formats where speculative projects are assessed without bias.

• Workshops where sensitive notes are anonymized before group discussion.

AnonBio opens new possibilities for trust, collaboration, and creativity in biodesign.

FAQs about AnonBio

Q: Does AnonBio store or transmit my data?
A: No. Everything runs locally in your browser.

Q: Can I customize the anonymization rules?
A: Yes, with the Foundational Tier you receive the full source code to adapt for your lab or studio.

Q: Will anonymization remove too much detail?
A: No. It replaces sensitive elements with placeholders, preserving design intent while hiding raw recipes.

Q: How does this fit into biodesign practice?
A: AnonBio acts as a lightweight safety step, encouraging more open feedback and dialogue without risking sensitive details.

Summary

AnonBio is a browser-based anonymization tool for biodesigners that protects sensitive details while enabling feedback and collaboration. By scrubbing identities, softening technical data, and providing safe previews, it creates a ritual of safe sharing for students, researchers, and founders.

👉 Try it free, or join the Foundational Tier to access source code, extended methods, and integration insights for biodesign practice.

Dear {{first_name | reader}},

We’re pleased to share an early release of our Mindtrek 2025 conference paper, available exclusively to Foundational members ahead of its official presentation in October. The paper reports on the first systematic accessibility audit of widely used molecular design tools — AlphaFold 3, ColabFold, ESMFold, and PyMOL — evaluated under the EU Accessibility Act (EN 301 549 / WCAG 2.1) using both sighted and screen-reader testing.

For students, the paper demonstrates how accessibility testing can be framed as a rigorous research contribution in the context of biodesign and protein engineering. For educators, it offers a structured, repeatable methodology that can be integrated into coursework, linking Human–Computer Interaction, accessibility, and molecular design practice. More broadly, the study highlights how inclusive design principles apply to the very tools that underpin biodesign education and research.

Read the Preprint (Foundational Access Only).

Best wishes,

Raphael, Biodesign Academy

Photo by Mehdi Mirzaie on Unsplash

What Our Accessibility Audit Reveals About Molecular Design Tools

Introduction

This early-release paper from Mindtrek 2025 presents the first systematic accessibility audit of major molecular design tools — AlphaFold 3, ColabFold, ESMFold, and PyMOL. The evaluation applies the EU Accessibility Act (EN 301 549 / WCAG 2.1), using both sighted testing and screen-reader-based assessment.

The research demonstrates how accessibility evaluation can serve as a rigorous scientific contribution in the fields of biodesign and protein engineering.

Why Accessibility in Molecular Design Tools Matters

• Molecular design platforms are central to biodesign education and research.

• If these tools are inaccessible, students, researchers, and educators with disabilities face barriers to participation.

• Inclusive design is not optional; it is a requirement under EU law and a cornerstone of ethical scientific practice.

Key Findings from the Accessibility Audit

1. AlphaFold 3

   • Strengths: High accuracy in structural predictions.

   • Limitations: Inconsistent keyboard navigation and missing alt text in visual outputs.

2. ColabFold

   • Strengths: Integration with Google Colab enables wide access.

   • Limitations: Screen-reader compatibility issues with dynamic rendering.

3. ESMFold

   • Strengths: Efficient computation of large-scale protein structures.

   • Limitations: Poor semantic labeling of interface elements for assistive technologies.

4. PyMOL

   • Strengths: Widely adopted visualization environment.

   • Limitations: Heavy reliance on mouse-based input, with limited alternative interaction models.

Educational and Research Implications

• For Students: Demonstrates how to frame accessibility testing as formal research, bridging biodesign, HCI, and inclusive design.

• For Educators: Provides a structured, repeatable methodology that can be integrated into coursework.

• For Researchers: Highlights gaps where inclusive design principles must be embedded into next-generation computational biology tools.

Broader Impact

• This study represents a first-of-its-kind accessibility evaluation of widely used molecular design tools.

• It reinforces the principle that accessibility is integral to scientific infrastructure, not an afterthought.

• The findings support the ongoing convergence of computational biology, accessibility standards, and biodesign education.

FAQs

Q1. Why was the EU Accessibility Act used as the framework?
The EU Accessibility Act (EN 301 549 / WCAG 2.1) provides legally recognized, rigorous criteria for accessibility compliance across digital tools, making it suitable for international research contexts.

Q2. How does this study contribute to biodesign education?
It provides a methodology students can replicate, connecting accessibility evaluation to protein engineering and computational design coursework.

Q3. What are the next steps for improving accessibility?

• Collaboration with tool developers to address identified issues.

• Expanding audits to additional molecular design platforms.

• Establishing accessibility as a core benchmark in computational biology research.

Summary

The Mindtrek 2025 paper establishes accessibility auditing as a research frontier in molecular design, demonstrating its relevance to students, educators, and biodesign researchers. By assessing AlphaFold 3, ColabFold, ESMFold, and PyMOL, the study shows how inclusive design principles must shape the future of computational biology tools.

👉 Read the Preprint (Foundational Access Only)