Founders & Startups May 2026 7 min read The Contrarian Case

The Companies That Do Not Fake It Are Losing the Race.

Engineer.AI had 1,000 engineers in India building what it claimed AI was building. Amazon's "Just Walk Out" had 1,000 humans in India watching your camera feed and adding items to your cart. The companies that honestly build the technology and admit its limits... those are the ones falling behind. This has to stop at the due diligence stage.

1000s

GenAI companies... the majority are wrappers.

Wrappers of Gemini. Wrappers of ChatGPT. Some of the Unicorns and Decacorns in the space are, on close inspection, also just wrappers. The underlying model is someone else's. The "AI" in the name is a marketing decision, not an engineering one.

"Fake it till you make it." A young chap told me this quite casually recently. I am not a supporter of the philosophy... but I am aware that a large number of startups operate by it. The question I find more interesting is: where does faking become fraud? And who is responsible for drawing the line?

I have two documented cases that I keep coming back to.

I am not sure why I still think about this. Maybe it does not matter.

Case Study 2019

Engineer.AI... "Getting an app built is as easy as ordering a pizza."

The claim: a proprietary AI system that could generate beautiful, functional applications at rapid pace. The pitch was compelling. The marketing was slick. Investors wrote cheques, though looking back it is baffling how little technical due diligence was actually done before the money moved. The company raised significant capital on the strength of this AI story.

What was discovered

It was not AI building the apps. It was thousands of low-cost engineers in India, manually building each application. The AI framing was the pitch; the reality was an offshore development shop with a better brand.

Case Study 2024

Amazon "Just Walk Out"... cashierless stores powered by AI.

The claim: computer vision and sensor fusion so sophisticated that you could walk into a store, pick up items, and walk out... no checkout, no cashier, no friction. The technology received enormous press coverage. It positioned Amazon as years ahead in retail AI.

What was reported

More than 1,000 humans in India were watching camera feeds of shoppers in real time, adding items to virtual carts as customers picked them up. The AI was the label on the box. The labour was inside it.

These are not small companies making exaggerated LinkedIn posts. Engineer.AI raised over $29 million. Amazon is one of the largest companies in human history. If they calculated that faking the technology was worth it... for valuations, for stock price, for competitive positioning... then the incentive structure has a serious problem.

I saw this at a much smaller scale in Pune last year. I was consulting for a logistics firm for three weeks, and they were paying Rs 85,000 a month for an "AI routing engine." It turned out to be a guy named Amit sitting in a back room with Google Maps and a laminated spreadsheet, manually texting drivers.

A large office floor with workers looking at retail store camera feeds on dual monitors — This is what "Just Walk Out technology powered by AI" looked like from the other side of the world.

The companies that genuinely build the technology... that are honest about its capabilities and limitations... they lose the race. They do not claim things they have not done; so obviously they will always appear to be behind the companies that only need to claim and do not need to deliver.

Actually, I realise this is a tangent.

I do not fully understand why companies like Amazon do it. It is not purely competitive... Amazon is not losing to a startup in cashierless retail. My best guess is that radical innovation announcements serve stock prices; that the AI story is for the market, not the customer. But I am not certain. What I am certain of is the outcome.

Companies that fake it

Attract more capital at higher valuations

Win press cycles and mindshare

Set expectations that honest competitors cannot match

Raise rounds before the truth surfaces

Companies that do not

Appear less impressive in pitch meetings

Lose deals to companies making bigger claims

Face tougher questioning because they volunteer limitations

Take longer to raise at lower valuations

We are a deep tech firm. We have raised a couple of rounds. In every due diligence process I have been through, the split has been roughly the same.

The due diligence problem

95% of VC due diligence is financial and legal. Almost none of it is technical.

Financial & Legal... 95%

Tech... 5%

For a company claiming to be tech-first... claiming that its AI is the product... this ratio is indefensible. A financial auditor can verify revenue. Very few VC firms employ people capable of verifying whether the AI being pitched actually exists in the form described. Engineer.AI raised $29M before anyone checked. Amazon ran a human call centre for years before anyone reported it. The gap in technical due diligence is not a minor oversight. It is the mechanism by which faking is rewarded.

I am not saying every company that raises capital is dishonest. Most are not. But the ones that are... the ones running a performance of AI rather than building it... they benefit from a system that does not look closely enough. And the cost of that is paid by the companies that do the work quietly, honestly, and without the theatre.

The fix is not complicated: VC firms investing in tech-first companies need to have, or hire, people who can actually evaluate the technology. Not just the deck. Not just the demo. The code, the architecture, the data pipelines, the model performance on held-out tests. If that becomes standard, the playing field changes immediately. Right now, the incentive points entirely the other way.