Every facility is a stack of documents that says the facility is complete. Complete is a paperwork state. The system is built to produce it. It is not the same as correct.

In this fourth episode:

- Why a handover package is signed off when the documents are present, not when the field is verified

- The incentive the contract builds in: payment is released against the document, so the document is what gets produced, and the as-built does not always catch up

- The sector-wide bill: the Get It Right Initiative (GIRI, getitright.uk.com) puts the direct cost of avoidable error in UK construction at about 5 percent of project value, and closer to 21 percent once latent defects and unlogged indirect costs are counted. Around £21 billion ($29 billion, €24 billion) a year. Roughly seven times the industry's annual profit

- The question for this week: of the data you are about to operate on, where would a wrong record hurt the most? Start there

Still Dark, the companion newsletter, publishes every Thursday on LinkedIn and at vistergy.com, on the gap between built and working. This episode is the operator's view of one place that gap hides: inside data that everyone signed off and no one verified.

Subscribe on Apple Podcasts, Spotify, or wherever you listen.

Host: David Ayeni | Permit to Operate.

---

Full transcript

It's the morning of commissioning. The engineer walks to panel 4B. The drawing says the breaker is there. The supplier signed it off in March. He opens a cabinet.

It's empty.

Welcome to Permit to Operate. I'm David Ayeni.

Every facility is a stack of documents that says the facility is complete. Complete is a paperwork state. The system is built to produce it. It isn't the same as being correct. The data may say the asset is there, but the field is where you find out the reality. That's where the rubber hits the road. This episode isn't about people being careless. It's about reality. It's about a system that pays for the document and never checks the wall, and what that costs you when you discover it late.

A handover package is signed off when the documents are present, not when the field is verified. Complete is a paperwork state. It doesn't mean the paperwork is true. It just means that, for example, a supplier marks an asset complete to hit a milestone payment. That's the incentive the contract typically builds in. In my experience, payment is released against the document, so the document is what gets produced. The as-built doesn't always catch up, and the system never asked anyone to walk the system down.

The bill is the real thing, sector wide. The Get It Right Initiative, or GIRI, puts the direct cost of avoidable error in UK construction at about 5 percent of project value. Once you count the latent defects and the indirect costs that nobody really logs, GIRI's own research puts it closer to 21 percent. That's around £21 billion a year. By GIRI's reckoning, that's roughly seven times the industry's annual profit.

The number you can see understates the number that is real. That's what this episode is all about. This whole episode shares that statistic.

The handover that passed every checklist can be wrong about every asset. Completeness is a count, to be fair. Correctness is the truth about the count that it's supposed to stand for. No one has the time to verify every record against the field. Who has the time? That isn't the bar, and it never was. So the question is, if the data you're about to operate on isn't correct, where would a wrong record hurt the most? Maybe we start there. Because that gap is the difference between two records. One you can operate on, and one you're about to find out in a hard way. Usually at the worst possible time.

Still Dark continues every Thursday, on the gap between built and working. This is the operator's view of one place that gap hides: inside data that everyone signed off and no one could verify.

I'm David Ayeni.

Permit to Operate.

You cannot hire your way out of the handover gap. The specialists who turn millions of construction documents into something an operations team can use are numbered in the dozens, not the thousands. The build pipeline is exponential. The talent pool is not.

In this third episode:

- Why the US construction industry entered 2026 short by around 439,000 workers, and why commissioning is the thinnest talent pocket in the entire data centre stack

- Why cross-training a senior engineer buys six months at the margin, not a multiplier

- Why a codified checklist makes a team consistent, but consistency is not the same as competence

- The sidestep: redesigning the work so data is structured as it is generated, not remediated eighteen months later

Still Dark, the companion newsletter, publishes every Thursday on LinkedIn and at vistergy.com. The recent edition, "No AI signs the permit," is the macro view: the signature is the bottleneck no model crosses. This episode is the operator's response to the workforce consequence underneath it.

Subscribe on Apple Podcasts, Spotify, or wherever you listen.

Host: David Ayeni | Permit to Operate.

---

Full transcript

A delivery director on a one gigawatt project gets the slip notice on Tuesday. She calls every specialist firm she knows on Wednesday. All of them are booked through 2027. Thursday morning, she sits down with her own team.

The team that exists is the team that ships.

Welcome to Permit to Operate. I'm David Ayeni.

A recent Still Dark made the structural argument. The gap between built and working is widening. Every coordinating function is bounded by human time. The demand is not. The signature, least of all. This episode is the operator's response to that argument. When you cannot hire your way out of the gap, what do you do? There are three answers in the field right now. None of them are about hiring more people.

The US construction industry entered 2026 short by around 439,000 workers. Most of them are skilled trades. Electricians, pipefitters, mechanical fitters. Inside that shortage, one role is the thinnest. Commissioning is the thinnest and most fragile talent pocket in the entire data centre stack. Handover at industrial scale needs people who can read 30,000 documents in three formats, structure them against 25 conflicting standards, and hold the design intent in their head while they do it. These specialists exist. Mostly self-trained, mostly in their forties and fifties, mostly already booked on a hyperscale project. Number them in the dozens, not the thousands. The market for their time is sized by the build pipeline, and the pipeline is exponential. Nobody has ten years to grow a new specialist when the next facility commissions in eighteen months.

Take a senior mechanical or electrical engineer, six months out of role. Train them on standards harmonisation, document extraction tooling, and commissioning workflow. The maths does not solve the gap. Cross-training buys six months at the margin, not a multiplier. Useful where it lands. Not the answer.

A codified handover checklist gives a junior team something to follow. It reduces the obvious mistakes and standardises the obvious answers. The non-obvious calls still need a specialist. When one standard contradicts another. When the model says one thing and the field says another. When the supplier marks an asset complete and the as-built shows it was never installed. The output of a checklist is consistent. It is not necessarily correct. Consistency is not the same as competence.

The operators who scale are not hiring more specialists. They are redesigning the work so the specialist's job arrives smaller. Every milestone in construction leaves a structured artefact. Each artefact is ingested as it is generated, not after the fact. When commissioning starts, the data is already structured. The specialist's job is not to process eighteen months of legacy. It is to verify the last fortnight of work. This is not theoretical. It exists at scale at one or two sites today. Most operators are not yet ready to budget for the rebuild. The ones who are will ship on time. The ones who are not will keep calling specialist firms and getting the same answer.

You cannot out-hire the gap. The handover specialist field is small. It does not scale with the build pipeline. The operators who scale are not the ones with the biggest team. They are the ones who redesigned the work before they ran out of people.

The question for this week is this: what part of your handover process exists because of who works on it? And what would it look like if you assumed they were not coming? That answer is going to define which projects ship on time over the next decade. Not the ones with the largest budget. The ones with the smallest dependency on people who are already on someone else's site.

Still Dark continues every Thursday. The recent edition, "No AI signs the permit," is the macro view: the signature is the bottleneck no model crosses. This is the operator's view of what you do underneath it.

I'm David Ayeni. Permit to Operate.

In this second episode:

- Why the connection queue is mostly fiction, and why ERCOT in Texas and PJM in the US East Coast face the same pattern

- Three years is fine for a power plant. Impossible for a one gigawatt AI factory that needs to switch on in eighteen months.

- "Behind the meter" is a verb now: Stargate Abilene's 1.2 GW dedicated power plant, Constellation reopening Three Mile Island for Microsoft, Amazon and Talen at Susquehanna

- What this means structurally for grid planners, regulators, developers, and everyone left in the queue

Still Dark, the companion newsletter, publishes every Thursday on LinkedIn and at vistergy.com. Today's edition, "The grid was not built for what is coming," is the macro view. This episode is the operator's response.

Subscribe on Apple Podcasts, Spotify, or wherever you listen.

Host: David Ayeni | Permit to Operate.

---

Full transcript

Britain's grid has seven hundred gigawatts of generation waiting to connect. Roughly twice what the country actually consumes. Most of those applications will never be built. The hyperscalers know this. Their answer is to stop waiting.

Welcome to Permit to Operate. I'm David Ayeni.

Last week's Still Dark made the macro argument: the grid was not built for what is coming. Today's episode is the operator's response. What do you do when the grid takes three years to give you a connection? When you have a one gigawatt facility that needs to switch on in eighteen months? There are two answers in the field right now. Wait. Or sidestep entirely. This episode is about the sidestep.

Britain's connection queue is seven hundred gigawatts. The country consumes three hundred and fifty at peak. Twice the national load is sitting in a queue. Most applications are speculative. Generators bidding for capacity they may or may not build. Developers locking in connection points before they have site, finance, or planning. The result: real projects cannot tell which applications ahead of them are real. Everyone is waiting in a queue that is mostly not moving. This is not a uniquely British problem. ERCOT in Texas. PJM on the US East Coast. Same pattern, different acronyms.

A typical connection request in Britain takes about three years from application to energisation. Some take longer. Three years is fine for a power plant or a heavy industrial site. The build cycle matches. Three years is impossible for a one gigawatt AI factory that needs to be online in eighteen months. The mismatch is structural. The grid was sized for a different economy.

Behind the meter used to mean a small-scale generator owned by a single facility. Solar panels on a warehouse roof. A backup diesel generator. Now it means a one point two gigawatt power plant built directly next to the data centre it serves. Stargate Abilene has dedicated power. It is not connected to the grid for primary load. It also means nuclear power purchase agreements, or PPAs. Constellation reopening Three Mile Island for Microsoft. Amazon and Talen at Susquehanna. The hyperscaler signs the entire output of a nuclear unit. Twenty years. Behind the meter is no longer a workaround. It is the design choice.

For grid planners, the largest loads in the country may never appear on their grid. They lose visibility into where the demand actually sits. For regulators, traditional rate-based economics break when the highest margin customers self-supply. Who pays for the grid? For developers, behind the meter changes the site selection calculus. Land near a closed nuclear facility is suddenly the most valuable real estate in the energy sector. For everyone else, the demand for grid connections that does land in the queue becomes more concentrated, not less. The seven hundred gigawatt number is a symptom, not the problem.

Behind the meter started as a workaround. It is becoming the architecture. When you cannot wait for the grid, you build a private one. When the cost of waiting is measured in millions of pounds or dollars per day, the calculus is straightforward.

The question for this week is this: when the largest energy customers in the world stop using the public grid, what does the public grid become? That question is going to define the next decade of energy policy. Not in the abstract. In the planning meetings happening right now. Behind doors most of us never see.

Still Dark continues every Thursday. Today's edition of the newsletter, "The grid was not built for what is coming," is the macro view. This is the operator's view.

I'm David Ayeni. Permit to Operate.

This happens everywhere. Nuclear plants in France. LNG terminals in Qatar. Data centres in Virginia. The industry calls it commissioning delay. I call it the gap.

In this first episode:

- Why handover is a phase, not a moment, and how 12 to 24 months of hidden work becomes revenue lost

- The intra-facility gap and the inter-sector gap. Built separately. Operated separately. Still dark separately.

- A gigawatt of AI training needs a gigawatt of baseload power. What that means across 565 facilities: nuclear, LNG, data centres.

- The question grid operators keep asking that no one has answered

Still Dark, the companion newsletter, publishes every Thursday on LinkedIn and at vistergy.com. If Still Dark diagnoses the problem, this podcast explores the solution.

Subscribe on Apple Podcasts, Spotify, or wherever you listen.

Host: David Ayeni | Permit to Operate.

---

Full transcript

The steel is up. The concrete is cured. The commissioning certificate is signed. Eighteen months later, the control room is still dark.

This happens everywhere. Nuclear plants in France. LNG terminals in Qatar. Data centres in Virginia. The industry is calling it commissioning delay. I call it the gap. The gap between when infrastructure is built and when it actually works. Welcome to Permit to Operate.

I'm David Ayeni. For the last few months, I've been writing a newsletter called Still Dark. Five editions so far, each one about a different version of the same problem. Facilities that are technically finished but aren't producing anything. Document handover that takes six months. Standards that contradict each other. Control systems that can't talk to the asset management platform. A thousand small failures that each delay the moment the lights come on.

What started as a series about handover has turned into something bigger. And that's what this first episode is about.

When we started the first edition, the focus was the entire facility: inside the plant, one terminal, one data centre. It felt abstract. But actually, handover isn't a moment, it's a phase, often after the construction is finished. There are micro handovers through the process. Every month too long is revenue lost, opportunity lost, because sometimes facilities have already sold futures. There are power purchase agreements in some cases. So the documents, the standards, the commissioning tests that occur before the handovers, these micro handovers, are really important. Each one can turn into a bottleneck that really exacerbates the problem. The pattern repeats across sectors. It's not a coincidence. A common architecture of failure.

What's going to change with this podcast is this: we know that the facility-level gap is real, but that's not the whole picture. There's a second gap, and it's bigger. The gap between sectors. Energy infrastructure and compute infrastructure are convergent. A gigawatt of AI training needs a gigawatt of baseload power. Reliable baseload power. But the energy sector doesn't know the AI sector really exists, until they start to commit a lot more money. And the AI sector thinks power shows up because it always has. Two industries building in parallel, not usually talking, but the owners know there's a challenge. That's the inter-sector gap.

I've been hearing a version of the same question from grid operators, power teams, regulators: which data centres are being built within baseload range of my reactor? No one has answered that question already. Not the operator, not the hyperscaler, not the regulator. 565 facilities across three sectors: nuclear, LNG, data centres. The pattern is the same. Built separately. Operated separately. Still dark separately.

The question, and the silence that follows it, is the reason this podcast exists. Standards. Handover. Human decisions. The handover phase no one tracks. The people who actually close the gap. Those are the people I'm going to be inviting onto this platform.

Some episodes will be solo like this one. Some will be conversations with operators, regulators, and people who actually sign the permits to operate. All of them will be short. All of them will be specific. All of them will be about the invisible phase between construction and operation. The phase where billions of pounds in capital sits waiting, especially when you don't have seasoned professionals on the project or on the teams. Where climate commitments go to die. Where the promise of new infrastructure meets the reality of getting it to work.

If the gap is a protagonist, then the question I want you thinking about this week is this: when you hear that a new facility has been finished, or done done, what do you think that actually means? Because in every sector I've looked at, finished and working are two very different things. And the difference between them is where the real story lives.

Still Dark continues every Thursday on LinkedIn and at vistergy.com. If that's the diagnosis, this conversation is the exploration.

Subscribe wherever you listen. I'm David Ayeni. Permit to Operate.