They promised the future. They lied.

We have a brand new series for you today. If you’re a dedicated reader, you may have noticed we kind of let this slip a couple of weeks ago with an incorrect subject line. Well, now you get to actually read it. Yay!

-MV

What’s in store:

They promised the future. They lied.
Your Weekly AI Roundup!
Poll of the week: AI’s dad energy?
We compared Claude Code with ChatGPT Codex!

Read Time: 6 minutes

COLD CASE

They promised an AI future. They lied.

It’s the summer of 2023. Hundreds of employees from the buzzy British startup Builder.ai have gathered in Vietnam to celebrate a $250M funding round. Bottles are popping. People are dancing. But…

The authorities are closing in.

Nothing kills a party faster than a money laundering summons.

Today we reopen the case on a company that promised an AI-powered future they couldn't deliver. The truth was more shocking than anyone could have imagined.

The Evidence File

To understand how we landed here, we’re going back to 2016.

Sachin Dev Duggal founded Engineer.ai, an “assembly line for software” that allowed anyone to build an app. No coding required.

Duggal’s favorite analogy was ordering a pizza. Just like adding toppings, Engineer.ai customers pick their features from a menu. Then AI cooks up a functional app, hot and ready to be served. 80% of every order was built with artificial intelligence. Human technicians just sprinkled on the finishing touches.

The start-up reported $24M in gross revenue and boasted major clients like the BBC, Virgin Airlines, and the San Francisco Giants. By 2019, they closed a Series A with SoftBank. The world was Duggal’s oyster.

…until the press started asking questions.

Investigative reporters at The Wall Street Journal published an exposé that cracked the story wide open. While Engineer.ai claimed that LLMs were the driving force behind every product, employee testimonies told another story - people – not algorithms – were doing the heavy lifting.

But rather than address the issues head-on, Duggal took the narrative into his own hands. He wiped the slate clean and relaunched the platform with a new name: Builder.ai. Duggal got lucky - in the AI boom, nobody was looking back - and by ignoring the accusations, they didn’t stick.

The Crime: Faking Business Structure + Inflating Revenue

Builder.ai continued on the same trajectory as its predecessor. Raising money and inflating numbers along the way.

By 2024, the startup was valued at $1.5B and claimed $220M in revenue - even though its books showed more like $50M. Just as The Wall Street Journal reported years earlier, AI was not responsible for the business’s output.

Instead, over 700 engineers out of Asia and Eastern Europe were manually creating every single app on the platform. There was nothing artificial about their intelligence.

What Builder.ai marketed as groundbreaking AI was just three engineers in a trench coat.

The Victims

When the numbers collapsed, real people went down with them. When the end came, it came fast. Hundreds of employees gathered from around the world for a remote town hall in May 2025.

New CEO Manpreet Ratia made a shocking announcement: Builder.ai had run out of money. The company was filing for bankruptcy, and they were all being let go.

Investors, including Microsoft and the Qatar Investment Authority, poured as much as $445 million into the company. Three guesses on whether you think they got their money back.

The Prime Suspect

As the self-proclaimed Chief Wizard, Duggal was better at illusion than execution. His motives seemed simple: money, status, and the intoxicating identity of being the guy who democratized software.

The generous interpretation is that Duggal believed in Builder.ai as a concept, but couldn't wait for the technology to catch up. So he leaned into the age-old philosophy of “fake it til you make it” - with cheap offshore labor, fabricated revenues, and a whole lot of charisma.

But when it comes to nine-figure funding rounds, that’s just... fraud.

The moral of this story: Companies that succeed in this new future won’t just have a compelling vision. They’ll have real technology to back it up.

Case Status

Duggal stepped down as CEO in 2025, but the damage was long done. Soon after, Builder.ai filed for insolvency, leaving behind roughly $5M in restricted funds and a mountain of unanswered questions.

The irony? What Duggal promised can now be achieved with tools like Claude Code and ChatGPT Codex. Duggal just wasn’t patient enough.

FYI: The details of this cold case are ongoing. All allegations are based on publicly available reporting.

Want to build an app with AI like Builder promised? Now you actually can. Get our guide on whether to use Claude Code or ChatGPT Codex for your project!

Claude Code vs ChatGPT Codex →

POLL OF THE WEEK

What sitcom dad energy does AI have?

Tries to help, makes it worse

Gives advice nobody asked for

Means well, fundamentally clueless

Other - tell us!

Full results will be at the bottom of tomorrow’s newsletter!

THIS WEEK IN AI

Jalapeño chips, a returning model, and the man who built the internet just handed in his badge

OpenAI named a chip after a spicy pepper. Fable 5 came back from government timeout. And the co-creator of TCP/IP retired at 83 while politely warning that AI agents need better manners.

Genuinely one of the more poetic weeks we've had.

Here's what mattered:

Google is watching its best AI researchers walk out for pre-IPO equity at Anthropic and OpenAI — including a Nobel Prize winner — because restricted stock options tied to a $4 trillion market cap simply cannot compete with a lottery ticket.
OpenAI built its own AI chip called Jalapeño with Broadcom, a task-specific ASIC designed to cut Nvidia dependency and move the company toward full vertical integration just in time for its IPO.
China's Zhipu AI released GLM-5.2, an open-weight model that matches top US systems in some cybersecurity bug-finding tasks — and because anyone can download and modify it, Washington now has a genuinely tricky problem on its hands.
Suno launched Spark, an artist support programme offering grants and mentorship to unsigned musicians — with a "Good Vibes Only" clause in the fine print that bans recipients from saying anything negative about the company.
OpenAI previewed GPT-5.6 as three models: Sol for power, Terra for balance, and Luna for speed — with stronger cybersecurity skills, extra safeguards, and pricing that suggests one AI brain now comes with backup dancers.
Ford rehired more than 300 veteran quality engineers after AI cameras failed to match human expertise in factory inspections — and promptly topped the JD Power quality rankings for the first time since 2010.
Astronomers built CIGaRS, an AI tool that analyses supernovae from images to measure dark energy up to four times more accurately — because the Vera Rubin Observatory is about to generate more cosmic data than any human team could ever sort through manually.
Anthropic restored access to Fable 5 after the US Department of Commerce lifted export controls, following new safety training that blocks the flagged jailbreak method in over 99% of cases and a promise to do more government paperwork before future launches.
Sam Altman called for a US-led global forum to set safety standards for advanced AI after the G7 summit in France — arguing that Silicon Valley should not be writing the rules, marking its own homework, and handing itself a gold star.
Vinton Cerf, co-creator of TCP/IP and the closest thing the internet has to a founding father, stepped down from Google after 20 years — leaving with a warning that AI agents will need formal communication standards, because plain English is famously too chaotic to be trusted.