We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Sora 2 is OpenAI’s flagship video generation model. It is the successor of Sora, which was launched in February 2024. While Sora met with mixed responses and usage due to the presence of Veo 2 and eventually Veo 3, Sora 2’s launch has generated a lot of interest.

Sora 2 is also available as a TikTok-style app on iOS.

I’ll leave you to judge how good the model is (and it can generate sound too). All the videos below are completely AI-generated:

• Sora 2 launch video

• The most popular character for Sora 2 videos has been none other than Sam Altman. Here’s Sam stealing paintings from Studio Ghibli.

Twitter tweet

• Anime fan-made edits are going to go out of hand now

Twitter tweet

• Someone tried generating Indian uncles with Sora 2, and it did really well

Twitter tweet

• Here’s Michael Jackson interacting with a bear. The physics are incredible.

Twitter tweet

• Here’s Rick and Morty. Basically, all animated shows can have unlimited episodes now.

Twitter tweet

• If you try older video styles, even human videos feel original.

Twitter tweet

Simply put, we’re staring at really interesting and scary times ahead!

Did you get the chance to try Sora 2 yet?

MindNote is an AI-powered notetaker built for students and B2B users.

Organizing thoughts often means juggling messy notes, scattered documents, and endless editing to create something clear and structured.

What if you could turn raw ideas into polished mind maps and summaries in seconds—letting AI organize and refine your thinking effortlessly?

Check out the lifetime deal for MindNote on Oncely offering unlimited access forever.

A free newsletter with the marketing ideas you need

The best marketing ideas come from marketers who live it.

That’s what this newsletter delivers.

The Marketing Millennials is a look inside what’s working right now for other marketers. No theory. No fluff. Just real insights and ideas you can actually use—from marketers who’ve been there, done that, and are sharing the playbook.

Every newsletter is written by Daniel Murray, a marketer obsessed with what goes into great marketing. Expect fresh takes, hot topics, and the kind of stuff you’ll want to steal for your next campaign.

Because marketing shouldn’t feel like guesswork. And you shouldn’t have to dig for the good stuff.

Subscribe Free

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Claude Sonnet 4.5 was launched yesterday with this bold claim:

Here are the trust-me-bro benchmarks for the latest offering by Anthropic:

Without wasting time talking about the bold claims by Anthropic and the PR team, let’s dive into some real reviews.

• “You’re absolute right!” hasn’t gone away.

Twitter tweet

• One thing is clear that the model does have a developer’s sense of humour.

Twitter tweet

• Here’s the result of a prompt where the model describes itself.

Twitter tweet

• Others were more realistic.

  

• Anthropic released a banger marketing video themselves where they asked all Claude models to clone Claude.ai

Twitter tweet

• Here’s an example of pagoda art generated by different models with the same prompt.

Twitter tweet

• Maybe it’s not there yet.

Twitter tweet

Did you try the model yet? What do you think?

Your Boss Will Think You’re an Ecom Genius

If you’re optimizing for growth, you need ecomm tactics that actually work. Not mushy strategies.

Go-to-Millions is the ecommerce growth newsletter from Ari Murray, packed with tactical insights, smart creative, and marketing that drives revenue.

Every issue is built for operators: clear, punchy, and grounded in what’s working, from product strategy to paid media to conversion lifts.

Subscribe for free and get your next growth unlock delivered weekly.

Sign up now

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 250! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Google’s AI company, Google DeepMind, has just unveiled Nano Banana, its latest image editing model, now live inside the Gemini app.

Nano Banana is designed to give everyday users Photoshop-level power with far less effort. It is now clearly the leading image editing AI model in the world!

The standout feature? Consistency. Whether you’re giving yourself a retro beehive haircut, dressing your dog in a tutu, or previewing new wallpaper in your room, Nano Banana keeps people (and pets) looking recognizably like themselves across edits.

But it’s not like this consistency is maintained forever - the more funky you get with your edit requests, the easier it is to lose consistency.

Here’s what you can do with it today:

• Outfit & location swaps - Try on costumes, professions, or decades while still looking like you.

• Photo blending - Combine images of you and your pet into a single scene.

• Multi-turn editing - Iteratively tweak rooms, objects, or outfits step by step.

• Style mixing - Transfer patterns or textures (like butterfly wings onto a dress).

All edits are watermarked - both visibly and invisibly with Google’s SynthID. This makes all images generated by Gemini identifiable.

Nano Banana is available to all Gemini users.

Twitter tweet

I tried it and shared the results on X. What are you generating with Nano Banana?

Top Publishers Hand-Selecting Amazon Brands to Promote this Holiday Season

This holiday season, top publishers are actively sourcing brands to include in their gift guides, newsletters, listicles, reviews, and more to drive high-intent shoppers straight to Amazon storefronts.

Here’s why it matters:

• Amazon brands are seeing a 5:1 conversion rate compared to their DTC site

• Millions of shoppers discover products through trusted publishers

Levanta is working directly with these publishers to introduce them to a small number of 7–9 figure brands.

If you qualify, your products could be featured in high-traffic placements that deliver predictable CAC and directly measurable sales.

See if your brand qualifies

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 251! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

August 2025 has started with a bang featuring mega launches by major AI companies. Here are the top launches summarized.

Claude Opus 4.1

Claude Opus 4.1 is Anthropic’s latest launch. An upgrade to the Opus 4, it is now their state-of-the-art coding model with a score of 74.5% on the SWE-bench Verified benchmark.

Opus 4.1 is now available to paid Claude users and in Claude Code. It’s also on the API, Amazon Bedrock, and Google Cloud’s Vertex AI. Pricing is the same as Opus 4.

Genie 3

Genie 3 is a groundbreaking launch by Google DeepMind. It is a a general-purpose world model that can generate interactive environments. Given a text prompt, Genie 3 can generate dynamic worlds that you can navigate in real-time at 24 frames per second, retaining consistency for a few minutes at a resolution of 720p.

This is the beginning of many future models and physical tools that will create agentic AI-based, realistic and interactive environments, which could become the gold standard of VR and AR.

gpt-oss

Finally, something ‘open’ from OpenAI. The company released gpt-oss-120b and gpt-oss-20b — two open-weight language models that deliver strong real-world performance at low cost. This is their first open-weight release since GPT-2.

The gpt-oss-120b model achieves near-parity with OpenAI o4-mini on core reasoning benchmarks. The gpt-oss-20b model delivers similar results to OpenAI o3‑mini on common benchmarks.

Eleven Music

Eleven Music is ElevenLabs’ new AI music generator, enabling creation of studio-grade tracks from natural language prompts across genres, with and without vocals, and in multiple languages.

Eleven Music is cleared for nearly all commercial uses, from film and television to podcasts and social media videos, and from advertisements to gaming.

GPT-5

Perhaps one of the most awaited AI launches, OpenAI has launched its new flagship model — GPT-5. GPT‑5 is a unified system with a smart, efficient model that answers most questions, a deeper reasoning model (GPT‑5 thinking) for harder problems, and a real‑time router that quickly decides which to use based on conversation type, complexity, tool needs, and your explicit intent.

GPT‑5 is available to all users, with Plus subscribers getting more usage, and Pro subscribers getting access to GPT‑5 pro, a version with extended reasoning for even more comprehensive and accurate answers.

Clearly, the best OpenAI model, it has disappointed many users as they expected GPT-5 to be a major step towards AGI, but the model is more of an incremental update to the previous OpenAI models.

Go From Idea to Video in Minutes—Not Hours

Creating content daily shouldn’t feel like a full-time job.

Syllaby.io helps you generate faceless videos in just minutes—no editing, no filming, and no burnout.

✅ Auto-generate engaging short or long-form scripts
✅ Add captions, voiceovers, B-roll & character-consistent avatars
✅ Schedule and publish to TikTok, YouTube, Reels, and more

Whether you're a solopreneur or agency, Syllaby gives you everything you need to scale your content—fast.

Get 7-days free!

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 251! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Years before AWS even existed, there was a company that ran Facebook, Twitter, LinkedIn, and more on the cloud. It was also the owner of Node.js. It invented containers before Docker was a thing. But then it disappeared.

This is the story of Joyent.

Started in 2004, the name Joyent was coined by David Paul Young.

In November 2005, Joyent merged with TextDrive. Young became the chief executive of the merged company, while TextDrive CEO Dean Allen, a resident of France, became president and director of Joyent Europe.

Jason Hoffman (from TextDrive), serving as the merged company's chief technical officer, spearheaded the move from TextDrive's initial focus on application hosting to massively distributed systems, leading to a focus on cloud computing software and services to service providers.

What followed was a visionary cloud company that had a jaw-dropping list of clients - Twitter, LinkedIn, and Facebook. Joyent was literally everywhere but invisible.

Joyent built their own operating system called SmartOS which used the concept of container virtualization. This was 8 years before Docker even existed!

This was a company ahead of the curve. One Joyent server could easily handle what multiple Amazon EC2 instances couldn’t.

Unfortunately, as brilliant as the team at Joyent was, they were all Sun refugees and hence built using Solaris while the developers of the world wanted Linux.

Things didn’t immediately go downhill in all aspects. While Joyent began losing ground to massive investments in the cloud by Amazon, they brought in Ryan Dahl, the creator of Node.js, and became its steward.

Node.js is today used by NASA, Netflix, PayPal and many more leading companies.

But Jeff Bezos had seen the cloud as the next big thing. AWS kept slashing prices. They may not have had the best product but they had a cheaper and easier-to-adopt product.

Joyent tried to match Amazon’s pricing but they could not match the DX due to their reluctance to leave Solaris.

By 2010, AWS became synonymous with ‘cloud,’ and Joyent was disappearing already.

In 2016, Joyent was acquired by Samsung for $125 million. Pennies compared to the then-valuation of AWS at $100 billion. Even Docker was a billion-dollar startup by then.

Things worsened under Samsung and Joyent completely shut down their public cloud in 2019.

The lesson? Being the best doesn’t guarantee that you’ll stay at the top. You need to adapt to the changing world around you.

Training cutting edge AI? Unlock the data advantage today.

If you’re building or fine-tuning generative AI models, this guide is your shortcut to smarter AI model training. Learn how Shutterstock’s multimodal datasets—grounded in measurable user behavior—can help you reduce legal risk, boost creative diversity, and improve model reliability.

Inside, you’ll uncover why scraped data and aesthetic proxies often fall short—and how to use clustering methods and semantic evaluation to refine your dataset and your outputs. Designed for AI leaders, product teams, and ML engineers, this guide walks through how to identify refinement-worthy data, align with generative preferences, and validate progress with confidence.

Whether you're optimizing alignment, output quality, or time-to-value, this playbook gives you a data advantage. Download the guide and train your models with data built for performance.

Download the guide

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 251! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

If you haven’t heard of it yet, OpenAI launched the ChatGPT Agent mode last week. This gives ChatGPT the ability to use its own virtual computer and complete tasks for you from start to finish.

Here is the launch video:

ChatGPT already has a Deep Research mode which can go through hundreds of websites finding relevant text to answer your queries.

In January, OpenAI had also launched something called ‘Operator’, an agent with access to your browser.

I haven’t tried it, but Agent feels like a combination of both these tools (the same was mentioned in the launch video too).

If you give a long query to Agent (like completely planning and booking a vacation), it spins up a VM, makes use of Deep Research, its browser GUI, and a terminal to accomplish the task.

OpenAI also warned that just like any other user on the internet, Agent is not safe from attacks. Hackers might be able to persuade the agent to enter sensitive information on fake websites. The team encourages users not to let the agent enter sensitive information; rather, user should ‘take over’ and do sensitive tasks themselves.

Early Feedback

David Gewirtz from ZDNet used 25 out of the available 400 queries to Pro users. He wasn’t satisfied with the Agent. Although the Agent understood the problem very well, it failed to accomplish the task 7/8 times.

Did you get the chance to try ChatGPT Agent? Plus users get 40 queries a month and Pro users get 400 queries a month.

Finally, a powerful CRM—made simple.

Attio is the AI-native CRM built to scale your company from seed stage to category leader. Powerful, flexible, and intuitive to use, Attio is the CRM for the next-generation of teams.

Sync your email and calendar, and Attio instantly builds your CRM—enriching every company, contact, and interaction with actionable insights in seconds.

With Attio, AI isn’t just a feature—it’s the foundation.

• Instantly find and route leads with research agents

• Get real-time AI insights during customer conversations

• Build AI automations for your most complex workflows

• Join fast growing teams like Flatfile, Replicate, Modal, and more.

Start for free today.

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

It is unclear who coined the term “context engineering” but the concept has been in existence for decades and has seen significant implementation in the last couple of years. All AI companies, without exception, have been working on context engineering, whether they officially use the term or not.

What is context engineering?

Before I even give my definition of context engineering, let me disappoint you a bit.

All prompts are context and all context is a prompt.

This is true. The words are interchangeable. People might use them to mean different things but technically speaking, every prompt is a blob of text sent to an LLM. If you attach files or a knowledge base to your prompt as ‘context’, it still is, at the end, a blob of text or a ‘prompt’.

For a model, it is all a string of text. It doesn’t matter whether you call it ‘prompt’ or ‘context’ or input or something else.

So, does that mean ‘prompt engineering’ and ‘context engineering’ are the same?

You can take whatever definition you want or call them the same if you want to but here’s how I draw the lines:

• Prompt Engineering: The process of crafting prompts to get the desired output from an LLM. Here, ‘prompts’ refer to the most immediate user-crafted instructions given to an LLM.

• Context Engineering: The designing of systems to organize context optimally for an LLM to solve a problem better. Here, ‘context’ refers to all the information given to an LLM including the system prompt, the user prompt, history, knowledge base, RAG, and more.

The output of context engineering is again a prompt. Hence I proposed my initial premise that the terms can be used interchangeably.

But to simplify, the general trends have gone this way -

• The most immediate user input to an LLM is a ‘prompt’

• The entire information given to an LLM in a call is the ‘context’

Therefore, the designing of the prompt and the context are two different things.

Context engineering is emerging as a much broader field that involves not only entering a well-structured prompt by the user but also giving the right information in the right size to an LLM to get the best output.

Why is it suddenly trending?

Did you find this article interesting? Read the complete version on my Medium.

Finally, a powerful CRM—made simple.

Attio is the AI-native CRM built to scale your company from seed stage to category leader. Powerful, flexible, and intuitive to use, Attio is the CRM for the next-generation of teams.

Sync your email and calendar, and Attio instantly builds your CRM—enriching every company, contact, and interaction with actionable insights in seconds.

With Attio, AI isn’t just a feature—it’s the foundation.

• Instantly find and route leads with research agents

• Get real-time AI insights during customer conversations

• Build AI automations for your most complex workflows

• Join fast growing teams like Flatfile, Replicate, Modal, and more.

Start for free today.

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 254! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

On June 12, Discord, Spotify, Snapchat, ChatGPT, Gmail, Google Drive, Google Maps, and numerous other services all went down together. Media houses were quick to describe this as an “internet outage,” with some pointing at more realistic possibilities and others speculating a massive cyberattack.

We know the reason now. Google Cloud Platform (GCP) went down for some time which resulted in multiple services using GCP going down including some Cloudfare services which are also used by several websites which may not directly depend on GCP. This made it look like a complete internet outage instead of some fault with just Google’s cloud offering.

This is yet another example of how much power cloud providers hold over the internet.

The outage across services lasted for about 2 hours - some recovered earlier and some went down late. This is a huge blemish for Google. It has been lagging in the cloud market share for years with Amazon Web Services (AWS) and Microsoft Azure much ahead of it.

From a technical perspective, this is what happened - the bug (originally a feature) in one of the GCP API services that caused the crash was originally committed into production on May 29th but it needed a policy change to get triggered. This policy change was not done and the bug (aka feature) lay dormant for a while.

On June 12th, in a separate change, a policy change was made which triggered the bug. It took the engineers around 40 minutes to identify the root cause and initiate the rollback to restore normalcy. Eventually after about 4 hours, operations were completely restored across all services.

GCP has around 11-12% market share in the world. AWS has around 29% and Azure holds about 22%.

The GCP outage is quite similar to the Crowdstrike incident that shook almost all Windows systems across the world.

Google “deeply apologized” for the incident. I wonder if any company tried suing Google for violating the GCP SLA which claims to provide 99.99% uptime.

Modernize your marketing with AdQuick

AdQuick unlocks the benefits of Out Of Home (OOH) advertising in a way no one else has. Approaching the problem with eyes to performance, created for marketers with the engineering excellence you’ve come to expect for the internet.

Marketers agree OOH is one of the best ways for building brand awareness, reaching new customers, and reinforcing your brand message. It’s just been difficult to scale. But with AdQuick, you can easily plan, deploy and measure campaigns just as easily as digital ads, making them a no-brainer to add to your team’s toolbox.

Learn more now.

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 254! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

The heading might sound a bit clickbait-y but Apple has gambled and gone against the wave of recent AI-filled tech demos from companies like Microsoft and Google. Just to prove my point before I summarize the major announcements, the word ‘Siri’ was used only 2 times in the 90-minute WWDC keynote address.

At the same time, it’s not that Apple is rejecting AI. As of now, it looks like Apple clearly knows that it is lagging behind in cutting-edge model training and development. Apple has shifted its focus on integrating Apple Intelligence subtly into various components of its ecosystem without explicitly mentioning it every time. Apple has modified its messaging the way you’d expect everyone to once the AI buzzword is stale. The question remains on the quality of Apple Intelligence and how it stands against the competition. More announcements are expected in 2026.

Here are the major announcements at WWDC 2025:

• It’s the season of glass: For all the upcoming operating systems, Apple has introduced a refreshed design language called Liquid Glass. This was perhaps the most noticeable change in the event. The new interface has translucent elements throughout the UI. The initial reactions have been mixed. It’s interesting but I know Steve Jobs would have fired the design team.

• Default apps redesigned: Many apps like the Camera app, Phone app, and Safari have been redesigned. The Camera one caught me by surprise because it has been pretty much the same for years with no major complaints from users. Camera has been greatly simplified. Phone now has some good features copied from Android counterparts like Pixel.

• Live Translation and On-Screen Search: Just like Google Meet, Apple Intelligence will now support Live Translation on FaceTime and calls. Apple’s Visual Intelligence has now reached Android’s Circle-To-Search level of searching.

• iPad OS 26 Upgrades: This was by far the most positive announcement for me. iPad will now support a windowing system just like a Mac. This makes iPad much better for multitasking and closer to laptops.

• VisionOS 26 Upgrades: One interesting upgrade in the Vision OS 26 is to pin things in the space around you. For example, you want a wall clock on a particular wall of the room only - done, it’s pinned to the particular area. This positioning is retained on restarting the device too!

• Lastly, all OS versions are aligned now - it’s WWDC 2025; so iOS 26, iPad OS 26, WatchOS 26, macOS Tahoe 26, and VisionOS 26.

Which announcement impressed you the most?

Science Stories You Can’t Get Anywhere Else

Feed your curiosity with Nautilus — a science newsletter for thinkers, seekers, and the endlessly curious. Each week, we bring you beautifully written stories at the intersection of science, philosophy, and culture. From the physics of time to the psychology of awe, our essays, interviews, and ideas dive beneath the surface and linger in the mind.

Join a global community of readers who believe that big questions deserve thoughtful answers. Whether you're a lifelong learner or just love a good mystery of the universe, Nautilus will challenge how you see the world — and maybe even yourself.

Sign up now and start thinking deeper.

Sign up today

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

I was scrolling through X as usual and came across a free ideas thread by Guillermo Rauch. I’m just resharing the ideas with my thoughts added (if you care about it).

Vibe Note Taking App

I have no interest in note-taking apps. Even the default Android Notes app is good enough for me. But millions of people use them, Notion being the most popular so it would definitely be cool to have one that can create notes and visualizations for you in any structure that you can convey with your words. It would still be neat but more flexible than any note-taking app in the world. Cool idea!

LLM-like Google Trends

This is interesting if you build it but you need data like a search engine for this. For Google all it would take is to add a LLM on top of Google Trends. Not worth it in my opinion. But hey what do I know? Maybe you can build it and earn enough before Google does this.

Cloud To Database Tool

Whatever data comes into your SaaS (let’s take an AI chatbot for example) gets organized into a neat data structure and stored in a DB for later use. Interesting idea - can be of great use for some businesses, worthless for others. B2B ideas like this make great money if you’re able to find the right customers.

Super Deep Research App

“Many times I fire up ChatGPT, Grok, and Perplexity in a bunch of tabs“ - I do the same thing every time too. This is a personal problem so I can definitely relate to it. I know there are multiple tools that offer different models in the same UI (haven’t really used any of them a lot). Don’t they provide cross-model Deep Research too? If not, they should!

Google Images with Better Search and Editing

This idea has been floating in the tech circles for quite a while now. The problem with this - you practically need to build a image search engine better than Google and then monetize while Google Images is free. And, if you somehow manage to do this, Google can destroy you in weeks by just copying your features. Not worth building imo. But again I can be wrong (I would love better image search too!).

Quick, hard-hitting business news.

Morning Brew was built on a simple idea: business news doesn’t have to be boring.

Today, it’s the fastest-growing newsletter in the country with over 4.2 million readers—thanks to a format that makes staying informed both easy and enjoyable.

Each morning, Morning Brew delivers the day’s biggest stories—from Wall Street to Silicon Valley and beyond—in bite-sized reads packed with facts, not fluff, and just enough wit to keep things interesting.

Try the newsletter for free and see why busy professionals are ditching jargon-heavy, traditional business media for a smarter, faster way to stay in the loop.

Join 4.2M+ Professionals Now

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Just a couple of days back, I wrote how Microsoft made some crazy announcements at the Microsoft Build 2025. Well, Google has one-upped them. In fact, it looks like Google has surpassed all the competition.

With the Google I/O 2025, the company has announced itself as a leader in all things AI while strengthening its already dominant core offerings like Google Search.

I went through all of the announcements and picked the most interesting ones for you:

• Gemini Pro 2.5 Deep Think and Gemini 2.5 Flash: New models from Google that push all other models down in the benchmark charts. Even the cheaper and faster Flash model is better than previous leader, OpenAI’s o3. I have been using Gemini Pro 2.5, it is really good.

• Stitch: A tool to design UIs with just prompts, provide code, or export them as Figma designs. I tried Stitch briefly and I didn’t find it too crazy.

• Jules: An async agentic coding agent. It’s basically slow so it’ll send you a notification when it’s done with a task. You can connect it to GitHub and it is available to everyone, no waitlist. Try it here.

• Imagen 4, Veo 3, and Flow: You must have seen this already. Google’s latest video generation model is amazing. Examples:

Twitter tweet

Google also launched a tool called Flow that aims at Hollywood-level video creation and editing. It is available exclusively too Google Pro and Google Ultra subscribers (more on the subscriptions below).

• Gemma 3n: Open models (that can be used and sold by anyone) that are almost as good as Claude Sonnet 3.7 and better than GPT-4.1.

• AI in Search: Now that the Gemini models are good, Google has decided to use it everywhere (fair enough). Soon, you’ll have the AI mode in Google Search (just like Perplexity) and complete operations done by AI in the Chrome browser (OpenAI’s Operator is an early example). You can also just click a picture or make a video with the Gemini app on your phone and ask anything about it that the internet can answer. This was inevitable.

• Project Beam: 2D video to 3D video. Video calling, remote meetings, movies, porn, OnlyFans, so many industries might change if this works!

• Real-time voice translation on Google Meet: Yeah, I think this one’s self-explanatory. Imagine how crazy a world without language barriers would be. Just like all the sci-fi movies where the entire universe speaks English.

• Google AI Ultra Subscription: Worst thing at the end. The final goal is always to make money and if you want to access all the above AI tools (and more), you’ll need to shell out $124.99 for the first three months and $249.99/month after that.

Some really awesome tools, some beyond expectations. Something that extracts an unavoidable reaction from me is the quality of the Veo 3 models which also supports sound along with realistic video generation. I can’t imagine the amount of slop that will be released on YouTube once this models becomes cheap enough. The dead internet theory is going to become real. There’s probably no way we can avoid it.

One opportunity I can see emerging is the probable increase in real-life events because you will not have too many real interactions on the internet anymore. Maybe with your friends and family but in a few years you’ll be more likely to encounter a friendly AI bot than a friendly stranger. You might spend hours arguing something about your favourite sport with an AI agent that sounds just like a human and not even realize it. Redditors are already experiencing this. Anyway, that’s a topic for another article.

Which announcement from Google excites you the most?

Achieve More With Limited Resources: 15 Small Budget Success Stories

Strategic marketing doesn't require enterprise-level spending. Our latest case studies reveal how 15 small brands achieved remarkable outcomes through creativity and smart resource allocation.

• Innovative tactics that delivered exceptional ROI with minimal investment

• Strategic approaches that helped small teams compete against industry giants

• Data-driven techniques for maximizing impact when resources are limited

Looking for ways to stretch your marketing budget further? These 15 mini case studies show exactly how these small brands made such big waves without breaking the bank.

Get the case studies for free here.

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

At the Microsoft Build, the annual event for developers, Microsoft has made several huge and unexpected announcements.

More than 50 announcements have been made already, I went through all of them, and here are some really important ones according to me:

• GitHub Copilot is Open Source: The GitHub Copilot extension in VS Code (which I regularly use) is now open-source under the MIT license. This means that you can now legally modify the Copilot code and sell it as a new product without being jailed.

•  Copilot is now part of GitHub: It is called “GitHub Copilot” but it was an AI-coding assistant that could be integrated with IDEs like VS Code, Visual Studio, IntelliJ. Not anymore. The Copilot is now part of GitHub - you can tell it to raise a PR, solve an issue, or run tests right from your IDE using the @github command or directly using the “Copilot” button on the GitHub website.

• NLWeb: Natural Language Web is an open project that aims to interact with the web using natural language. It is in its early stages; you can find the GitHub repo here.

• Windows ❌ LLM-OS ✅: You will not find it worded like this anywhere but Microsoft is not enabling MCPs across Windows apps. This will be available in a private developer preview with select partners in the coming months. Initially, the plan includes the MCP Registry for Windows and MCP Server for Windows.

• Publishing apps to Microsoft Store is free now: Starting from June 2025, individual developers can publish apps to the Microsoft for free. This makes Microsoft Store, the first such platform in the world.

• Microsoft Discovery: A platform to enable the use of AI agents for scientific discoveries. Microsoft Discovery is built on top of a powerful graph-based knowledge engine. It is working already! -

• Azure AI Foundry and Windows AI Foundry: Two different things with the same name (just like Copilot for Windows, Copilot for Teams, and GitHub Copilot).
  Started in March, Azure AI Foundry is a platform to design and manage AI apps and Agents with the power of Azure. Apart from the 10,000+ open-source AI models, Foundry also has models like Grok 3 by xAI, and models like Flux Pro 1.1 and Sora to be added soon.
  Windows AI Foundry is an evolution of the Windows Copilot Runtime. It offers a unified and reliable platform supporting the AI developer lifecycle from model selection, optimization, fine-tuning and deployment across client and cloud.

Which announcements caught your eye?

100 Genius Side Hustle Ideas

Don't wait. Sign up for The Hustle to unlock our side hustle database. Unlike generic "start a blog" advice, we've curated 100 actual business ideas with real earning potential, startup costs, and time requirements. Join 1.5M professionals getting smarter about business daily and launch your next money-making venture.

Get the guide

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

On the 4th of May, billionaire investor and the CEO of Berkshire Hathaway, Warren Buffett announced his surprise decision to step down from his position at the end of this year. From 2026, Greg Abel will lead the multinational conglomerate. Buffett will stay on as Berkshire’s board chair — but either way it’s the end of an era.

I will not discuss the economic impact of this decision in the USA and across the world. You might have a broad idea about this already. But I take this as an opportunity to make you more aware of what a great personality Warren Buffett is.

Born to Howard Buffett (stockbroker and US Congressman) and Leila Buffett (homemaker) in Omaha, Nebraska, on August 30, 1930, Warren had the entrepreneurial spirit even as a child. He used to sell Coca-Cola, magazines, and chewing gum to make money and bought his first stock at just 11 years of age. Gotta start early!

Buffett graduated with a degree in business administration from the University of Nebraska-Lincoln in 1951. A major influence on Buffett’s investment philosophy was Benjamin Graham, the father of value investing, who taught Buffett at the Columbia Business School for his Master of Science in Economics. It was here that Buffett developed his hallmark investment strategy - buy undervalued companies and hold long.

Buffett first worked at his father’s brokerage company, then with Graham at the Graham-Newman Corporation, and then at several investment firms for the next 5 years.

In 1956, Buffett returned to Omaha and began building his now famous portfolio. This is where the game began!

In 1962, Buffett began buying the shares of a relatively unknown and struggling textile company. Eventually, Buffett gained control of the company in 1965 and began using it as a holding vehicle for a list of profitable businesses. He transformed the textile company into a conglomerate. That company’s name? Berkshire Hathaway.

Between 1965 and 2023, the company's shareholder returns amounted to a compound annual growth rate (CAGR) of 19.8% compared to a 10.2% CAGR for the S&P 500. With Berkshire Hathaway, often compared to an investment fund, Buffett has cemented his legacy as one of the greatest investors of all time.

Buffett is one of the leading figures of “value investing” - a long-term strategy based on thorough research, patience, and a focus on intrinsic value rather than market speculation. His strategy has been proven correct a number of times with successful investments in companies like Coca-Cola, American Express, Apple, and more.

In 2008, he became the world’s richest person and is still the world’s 5th richest person in 2025. Despite such enormous wealth, Buffett is known for his humility and modesty. Buffett purchased a house in Omaha in 1958 for $31,500. To this day, Buffett lives in the same house.

His annual salary for Berkshire Hathaway has been a meager $100,000. In terms of salaried income, many software engineers make more than him!

Buffett has a clear plan for his billions after his death. He will be donating 99% of his fortune mostly through the Bill & Melinda Gates Foundation.

Buffett’s personal values and ethics have seeped into the Berkshire Hathaway culture. Over the years, his annual letters have become a goldmine of investing wisdom. His deep relationships with business partners like Charlie Munger have highlighted a strong human aspect in the world of investments, where people often get lost and lonely in the chase after money.

Since the 1960s, Buffett has been a cultural icon. Being rich and respected for over 65 years is an achievement that almost everyone in the world can only dream about. Buffett has lived that dream.

His annual meetings in Omaha, like the one where he announced his retirement, have become a pilgrimage for investors who come from across the world to listen to the words of wisdom by him and Munger. His speeches are often laced with humour, personal stories, and lessons about integrity in the finance world.

Through his glorious career, Warren Buffett has not only established himself as an immortal symbol of successful investments but also as a man of legendary wisdom, integrity, philanthropy, and humility. The impact of his career will resonate for generations to come.

From a young, curious boy, to the “Oracle of Omaha”, Warren Buffett’s life is proof of the power of education, integrity, and long-term thinking.

The world will struggle to find another man so humble and consistently successful throughout his life.

Trusted by 1,000,000+ readers across the world and political spectrum

Ground News lets you compare how left, center, and right-leaning outlets cover the same story, so you can easily analyze reporting and gain a well-rounded perspective on the issues that matter to you.

Built for the age of the algorithm, its Blindspot Feed shows stories underreported by the left or right, helping you break out of your bubble and challenge your worldview. Because the news you don’t see can shape your perception just as much as what you do see.

Find out why the Nobel Peace Center endorsed Ground News as “an excellent way to stay informed, avoid echo chambers, and expand your worldview.”

Start your free trial today

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 254! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Anima Playground is a UI-first platform that enables you to go from a Figma design to a fully functional app in seconds. Today, we are going to do just that. I will design a minimal non-functional world clock in Figma and by the end of this article we will have a fully functional world clock coded, pushed to GitHub, and deployed by Anima Playground.

Step 1: Make The Initial Design In Figma

For the original design, I used a ‘Desktop’ frame in Figma with an indigo background. A simple “World Clock” title to the top left, a PNG world map image with a lighter shade of blue in the background and a red line in the middle, and two time-boxes complete our design.

The first time-box shows the local time, time zone, and location. The red line is supposed to be a slider that points to the area whose time and time zone is supposed to be shown in the second time-box.

Of course, it’s just a static design in Figma. Let’s see if Anima Playground can understand all the logic involved here.

Step 2: Load Your Design Into Anima Playground

Just open animaapp.com and paste the link of the Figma design to convert it into React code in seconds.

I pasted my World Clock’s Figma link and here’s the initial result:

Uh oh! This isn’t really our design. Let’s see if Anima Playground is smart enough to fix the mistakes.

Step 3: Fix Initial Mistakes

Anima Playground had the right components coded but some wrong repetitions of images and text so I tried this prompt:

“The time boxes on the bottom look fine. But I want a single map which is the world map image in the middle. And no extra images or time boxes. The “World Clock” heading at the top is fine but I don’t need “World Clock” repeated too. Just one heading, one map with the line, and two time boxes.”

And voila, we have the image and time boxes in the right place:

I also changed the background to white and maintained the rounded corner style that the Playground chose for me.

Step 4: Add Some Logic

Next, I decided to make the slider red and movable across the image. Anima understood and implemented this in a single attempt!

Step 5: Show The Local Time

As per the initial design, I wanted the first box to show the local time, time zone, and location. This should be relatively easier to code because JavaScript already has built-in functions to get the local time and time zones.

The Playground initially got the location as “Calcutta” from Asia/Calcutta as my time zone but I told it to use the city’s correct name — Kolkata.

Step 6: Get Time From Map

Well, here’s the tricky part. If you ask Anima Playground or rather any AI agent to directly get locations from the silhouette image that I’ve used, it will not work. After all, it is just an image and not a programmatically-marked map component.

So, for the world clock to function i.e., show different locations, times, and time zones as we drag the slider, we need to create an illusion.

I used this prompt — “The image that I’ve used is a black silhouette of the world map. What we need now is to dynamically change the time, timezone, location in the second time box as the slider moves across the world map. Of course, we cannot extract actual locations from the map so we need to create the illusion of changing locations as we move the slide from left to right. Have at least one location for each timezone. Starting from -11 in Alaska to +12 in New Zealand I guess. You need to pick the right time zones for the right location.”

Bingo! Anima Playground created a map (the data structure, not the image) of cities with the right time zones and used it to arrange them across the world map. It wasn’t perfectly aligned but it did the job.

For some reason, Playground also changed the code of the first box and my location changed to Karachi, Pakistan.

Step 7: Apply Final Touches

My final fixes included instructing Anima to fix my location and changing the original image so that the slider mechanism was more accurate (also, because the original map did not have New Zealand and I felt bad for them).

Here’s the final product:

Don’t expect it to be completely accurate with the locations on the map but it’s a usable project if you want to convert your local time to some other time zone.

Here’s the world clock app deployed by Anima: World Clock Powered By Anima Playground.

If you want to take a look at the code, you can check it out on my GitHub: Source Code (pushed in a single-click by Anima!).

Conclusion

This entire process took me less than two hours and I had no strict plans before starting. Everything was reactive and conversational between me and Anima Playground. Converting your Figma designs to fully-functional apps is easier than ever before.

Try the new Anima Playground for free!

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 253! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

OpenAI has launched their latest flagship models - GPT-4.1 Nano, GPT-4.1 Mini, and GPT-4.1. These models outperform GPT-4o and GPT-4o Mini. These are the first set of models that support a 1 million context window.

GPT‑4.1 scores 54.6% on SWE-bench Verified, improving by 21.4% over GPT‑4o and 26.6% over GPT‑4.5—making it a leading model for coding.

GPT-4.1 is already available for free in coding assistants like Github Copilot and Windsurf.

The models also show gains in real-world tasks like frontend development, with evaluators preferring 4.1's web interfaces 80% of the time over GPT-4o.

The GPT‑4.1 family is exceptionally strong at image understanding, with GPT‑4.1 mini, in particular, representing a significant leap forward, often beating GPT‑4o on image benchmarks.

It is also rumoured that these models might be the precursor to OpenAI’s latest project, Agentic Software Engineer.

GPT‑4.1 is 26% less expensive than GPT‑4o for median queries, and GPT‑4.1 nano is OpenAI’s cheapest and fastest model ever. Long context requests are now offered at no additional cost beyond the standard per-token costs.

Here’s the official launch video:

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.

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 254! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Nintendo Switch 2 was unveiled last week but it hasn’t yet become available for sale in the US because Nintendo is still assessing the impact of tariffs imposed by the Trump administration and how they are affecting the global trade. Nintendo is not alone. Tech companies around the world have entered into a zone of uncertainty due to global trade tensions.

Markets around the world have turned red. Trump announced tariffs upwards of 30 percent on both China and Taiwan on Wednesday as part of an expansive lineup of tariffs targeting imports across the board.

All tech companies that depend on manufacturing and supply chains through China and Taiwan have been affected.

China will be hit with a 34% tariff, on top of the 20% tariffs levied on Beijing over the past two months. Taiwan faces a 32% import tax, while goods from the European Union (EU) will encounter a 20% tariff upon entry into the U.S.

Apple lost over $300 billion in the market as its supply chains were hit hard by the tariffs. About 80 percent of Apple products are made in China.

Apple has been exploring other centers for manufacturing like India and Vietnam. Unfortunately, Vietnam is also going to face one of the highest tariffs in the world - 46%. India will also face a 26% tariff.

Semiconductors are exempted from tariffs directly but here’s the thing - semiconductors are usually inside something that is covered by the tariffs like laptops, smartphones, etc.

A notable exception that will not be affected by American tariffs is Tesla. The electric vehicle firm produces all of its vehicles in the U.S. at factories in California and Texas. Although there are fears that companies like Tesla can be affected by retaliatory tariffs by other countries.

Apple, Microsoft, Amazon, Alphabet, Meta, Nvidia and Tesla — a group also known as the “Magnificent Seven” — collectively lost more than $1 trillion in Thursday trading on Wall Street.

The Trump administration has earlier expressed how they want to position the US as the leader in AI but recent measures have been counterproductive.

Trump last month also called on Congress to scrap the “horrible, horrible” CHIPS and Science Act, passed in 2022 under President Joe Biden. The CHIPS Act is a bipartisan law that authorized nearly $300 billion in funding to boost semiconductor research and manufacturing.

China is not backing off and has imposed retaliatory tariffs of 34% and the Chinese are showing any signs of slowing down. The country is allowing the Yuan to weaken to make their exports more attractive.

The problem is no one knows when these tariffs will end. There is no clear end goal announced by the US government.

Learn AI in 5 minutes a day

What’s the secret to staying ahead of the curve in the world of AI? Information. Luckily, you can join 1,000,000+ early adopters reading The Rundown AI — the free newsletter that makes you smarter on AI with just a 5-minute read per day.

Sign up to start learning.

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 255! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Some popular images generated using ChatGPT in ‘ghibli anime style’

Ghibli Studio is my favourite anime studio with some of the most beautiful movies ever. Some of my favourites are Princess Mononoke and Howl’s Moving Castle. Add them to your watchlist.

Over decades Ghibli’s distinctive art style has become instantly recognizable to anime-watchers around the world.

AI image generation has been around for more than two years now. You must have heard names like Midjourney, Stable Diffusion, and DALL E 2. Different AI image generator tools and models have been capable of generating ‘Ghibli-esque’ images since some time but what happened in the past week that has made Ghibli-styled AI images probably the most popular internet trend of 2025?

4o Image Generation

On March 25, Sam Altman posted this:

Twitter tweet

The image generation capabilities in GPT 4o was launched on 25th March with a livestream by OpenAI.

Around the 3:35 mark in the live stream replay below, the team clicks a selfie and converts the image into anime style:

I consider this as the origin of the current Ghibli images trend. People saw what they can do with their own images in the demo. I consider this the origin of the current Ghibli image trend. In the demo, people saw what they could do with their own images. This felt like the most personal part and it took off - “What if I convert my image to anime style?”

Sam Altman changed his X profile picture to an anime-styled one generated by ChatGPT fueling the trend further.

Ghibli Fever

The Ghibli trend has been noticeable especially on X and Instagram with profiles from different fields including media houses, sports pages, politicians, celebrities, and others who began posting images in anime style made with 4o.

People without ChatGPT Pro access also tried different models like Grok 3 and Gemini with varying results.

Some popular examples:

Twitter tweet

Twitter tweet

Twitter tweet

OpenAI Runs Out Of GPUs

Three days into the trend, OpenAI ran out of GPUs because there were so many image generation requests by users worldwide. GPUs are critical for AI model training and generation, and the trend's viral nature may have strained resources.

Twitter tweet

Is It Fair?

The AI art debate has been raging on for the past couple of years.

Is it fair to the artists whose copyrighted works OpenAI and other companies have allegedly used in training their models?

If GPT 4o and other models can generate such likenesses of the Ghibli-styled animations, is it not clear evidence that the hard work of Ghibli artists over decades has been fed into a model that can generate many more such works of varying quality?

If yes, don’t the original artists deserve royalties? But how would such an arrangement be ever established?

Copyright laws on AI have been ambiguous. How can you strike down something that is, in fact, not owned by anyone? AI-generated work is new, created by an LLM, and owned by no one.

This is where the debate stalls and AI companies are exploiting this loophole to the fullest. Try generating an image of Thanos or JD Vance or Elon Musk on Mars with any popular AI model and you will get quite a realistic result.

Businesses have now turned a blind eye to AI ethics. There’s no time to discuss ethics. No company wants to lag. We are talking hundreds of billions of dollars of investors’ money here - copyrights, IP, and ethics can all take a backseat for now.

Not Just Ghibli

Let’s not undermine GPT-4o’s image generation capabilities by just talking about anime images.

GPT-4o has also achieved character consistency just like the Gemini Flash Experimental models but better.

But, the biggest advantage is the accurate text generation by the model:

Example shared by OpenAI

Due to the character consistency, GPT-4o also allows multi-step generations. There are many other powerful features which can make ChatGPT a great tool for graphic designing. ChatGPT can make consistent ads, posters, YouTube thumbnails, logos, etc.

This also probably creates an environment of threat among some designers. My opinion is designers should celebrate and incorporate ChatGPT into their workflows. Use ChatGPT for both inspiration and some quick wins while freelancing.

New World Order

We have to embrace it. AI art and image generation are here to stay. You can abhor it as much as you want but it delivers both aesthetic and business value.

It also feels like a good idea to remind that existing artists and designers are going nowhere. Human art will still be valued as much as it was valued earlier if not more.

For small businesses who know how to use AI, visual editing has become a whole lot easier and cheaper. AI is turning out to be a great business enabler in this sense.

The editing features are only going to get better. The same will apply to video and audio editing in the coming few months or years. We will get movies made with AI tools in the coming years complete with character consistency. Individual artists will become movie makers.

Humanity will be enabled by stochastic parrots to unleash their creativity faster than ever before. Human creativity will be irreplaceable at least for the foreseeable future.

Your job called—it wants better business news

Welcome to Morning Brew—the world’s most engaging business newsletter. Seriously, we mean it.

Morning Brew’s daily email keeps professionals informed on the business news that matters, but with a twist—think jokes, pop culture, quick writeups, and anything that makes traditionally dull news actually enjoyable.

It’s 100% free—so why not give it a shot? And if you decide you’d rather stick with dry, long-winded business news, you can always unsubscribe.

Check it out

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 254! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Popular AI-Assisted Coding Tools

It all started with this.

Twitter tweet

Computer scientist and popular internet personality, Andrej Karpathy coined a new term for AI-dependent coding - “vibe coding” - giving up the control to AI while coding.

AI assistants such as Cursor, Windsurf, etc. are becoming increasingly popular. These AI-powered tools are capable code generators and can translate user prompts into working code in almost any programming language.

This makes developing projects much faster even for people with beginner knowledge of coding.

The Good

In the indie hacking (individual project-making tech people) world, vibe coding has got a lot of hype after a super-successful project by one of the richest in the community, Pieter Levels.

Late in February, Pieter posted about how he’s vibe-coded a flight simulator game with almost no prior game development knowledge.

Twitter tweet

Even Elon Musk reposted the game and things took off. Pieter now sells planes and creative ad spots in the game and is now making $87K/month. He already had a huge audience and it became easier for him to get concurrent players on a game that is fairly simple and doesn’t involve any complex graphics, storyline, or physics.

This has made vibe coding a movement. Pieter himself is organizing a Vibe Coding Game Jam where people use AI to create games and the best games get rewarded. Karpathy and famous game developer, Tim Soret, are among the judges.

Popular AI coding tool, Bolt, has taken vibe coding to the next level with the world’s largest hackathon with prizes totaling more than $1 million.

Vibe coding is here to stay. It is backed by investors putting money into AI-assistant companies which will keep improving their tools and keep organizing more events to encourage vibe coding further.

The Bad

The problem is many users ignore and will ignore some important words in Karpathy’s original post -

Andrej is an elite coder. If something goes wrong in his vibe-coded application, he can debug it. It’ll take him some time to understand the code that he’d blindly been accepting for some time but he’ll eventually get there.

If you’re an inexperienced coder and try what I like to call - “vibe debugging,” you are going to have an extremely hard time.

If you are not a coder, you can still try vibe coding and get good results in the beginning but once the AI tools start losing context and start injecting unintentional buggy code, your application is doomed.

What can be even worse in the future is when the vibe coding movement gains pace and interns with production access at big tech companies start introducing vibe-coded security vulnerabilities into widely used applications.

Imagine a situation where you lose all the money from your bank account because an intern logged your net banking password in plaintext in the browser.

With the age of vibe coding, the golden age for bug bounties and hacking might also arrive.

My Conclusion

Pure vibe coding must be restricted to fun projects that do not have any scope of leaking sensitive information like credit card numbers or addresses.

To be a vibe coder, you must first be a coder. You can vibe code without learning coding but then it is at your own and your users’ risk.

AI-assisted coding is here to stay. I use it actively too. It is the coder’s responsibility to understand the code that they are implementing.

The gold standard of business news

Morning Brew is transforming the way working professionals consume business news.

They skip the jargon and lengthy stories, and instead serve up the news impacting your life and career with a hint of wit and humor. This way, you’ll actually enjoy reading the news—and the information sticks.

Best part? Morning Brew’s newsletter is completely free. Sign up in just 10 seconds and if you realize that you prefer long, dense, and boring business news—you can always go back to it.

Join 4.3 Million Readers Now

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 255! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

The engineers at Microsoft have done something that no one expected.

Under the project codenamed “Corsa”, the TypeScript team led by Anders Hejlsberg (the creator of C#, Turbo Pascal, Delphi, and TypeScript) has ported the TypeScript language from TypeScript to Golang compilation.

To put it simply, until now, TypeScript was written in TypeScript (surprise!). And, the problem with that is it is slow.

TypeScript does not have its own runtime. It is a superset of JavaScript. Every TypeScript application gets compiled to JavaScript and then runs on your browser or something like Node.js or Bun.

TypeScript is nowhere close to a low-level programming language to perform optimizations through pointers or multi-threading (for example).

The developers of TypeScript ported the entire language, i.e. converted line-by-line, into Golang.

With this, the Go compiler for TS is now 10x faster than the original TS compiler.

‘Native’ refers to the Go compilation in this image

Everything related to TS will now be 10x faster - the code compilation, the suggestions in your IDE, the error detection, the linting - everything.

What’s arguably even more incredible is that the team kept their egos aside and chose the best tool for this improvement.

The team did not choose C# which is Microsoft’s baby and could have been a great choice too. They also avoided other good and more hyped choices like Rust and Zig. They chose Go because after multiple studies, they found it as the best choice.

It might sound simple but in the corporate world, this is a bold choice. Someone stood up for what is right for performance (Go) rather than what could have been better for business (C#).

The current TypeScript version is 5.8 and the new native TypeScript (Go) will be released with TypeScript version 7.0. So this may be a few months or even more than an year away.

Here is the staging repository for typescript-go.

If you use TypeScript (or JavaScript), I strongly recommend you to watch this video by the OG Anders Hejlsberg himself -

Your job called—it wants better business news

Welcome to Morning Brew—the world’s most engaging business newsletter. Seriously, we mean it.

Morning Brew’s daily email keeps professionals informed on the business news that matters, but with a twist—think jokes, pop culture, quick writeups, and anything that makes traditionally dull news actually enjoyable.

It’s 100% free—so why not give it a shot? And if you decide you’d rather stick with dry, long-winded business news, you can always unsubscribe.

Check it out

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.

We are now a community of 254! Thank you❤️

This newsletter is free and I don’t use paid advertising. I completely rely on organic growth through users who like my content and share it.

So, if you like today’s edition, please take a moment to share this newsletter on social media or forward this email to someone you know.

If this email was forwarded to you, you can subscribe here.

If you want to create a newsletter with Beehiiv, you can sign up here.

Chinese startup Monica, owned by The Butterfly Effect, has launched a general-purpose multi-model AI agent called Manus.

What is Manus?

Unlike traditional AI LLM models which usually give a text-based reply and need multiple input prompts to achieve the desired result, Manus can perform a whole range of relevant tasks with a single prompt.

Whether you need to design and book your entire itinerary to Europe, analyze stocks and create interactive websites, or screen multiple resumes in a few minutes, Manus can do it all.

The Next DeepSeek Moment

The initial response to the official demo by the company has been positive. Some describe it as the next big AI launch since the launch of Deepseek R1 that shook the AI world.

The agent has outperformed leading AI models on the GAIA benchmark. Manus operates independently on the cloud. It has its own computer and access to the internet. It can also fetch data from social media websites.

More Hype Than Substance?

Early users have also reported performance issues. Some users have encountered huge delays and endless loops while performing simple tasks like boking a flight or ordering food.

Critics argue that exclusivity and marketing have fueled Manus’ popularity more than its actual capabilities. The developers of Manus have described the closed beta as a stress test.

Manus is currently invite-only. You can join the waitlist here.

Cut Through Noise with The Flyover!

The Flyover offers a refreshing alternative to traditional news.

We deliver quick-to-read, informative content across sports, business, tech, science, and more that cuts through the noise of mainstream media.

The Flyover's talented team of editors meticulously collects the day's most important news, ensuring you stay informed on top stories and equipped to win your day.

Join over 950,000 savvy readers and leaders who trust The Flyover to provide unbiased insights, sourced from hundreds of outlets!

Subscribe with One Click!

Did you like today’s newsletter? Feel free to reply to this mail.

This newsletter is free but you can support me here.

I’d be happy to connect on Medium, LinkedIn and X.