We Asked AI Insiders If The Industry Is a Scam

This week’s Visionary Voices sits down with Archie Chaudhury and Ram Shanmugam, co-founders of LayerLens — the AI evaluation company building the trust layer for the agentic era.

The AI industry has a problem it doesn’t like to talk about: nobody really knows which models actually work.

Benchmarks get gamed, demos polished and all of it dazzles investors. Meanwhile, companies pour billions into tools they can’t verify.

That’s the gap LayerLens is trying to close. Co-founded by AI researcher Archie Chaudhury and serial operator Ram Shanmugam, the company builds infrastructure for evaluating LLMs, testing agents, and letting enterprises create their own benchmarks. In a market high on promises and short on proof, they’re selling something unfashionable: evidence.

Is AI a Bubble?

AI is being compared to the 90’s dot-com bubble, and for good reason.

On January 27th, 2025 NVIDIA droppeds 17% in a single day. $600 billion gone, making it the largest single-day loss of market value by any company in the history of the American stock market.

Chaudhury didn’t flinch when we asked if this is a similar bubble to past ones.

“I certainly don’t think there’s like an FTX level thing that’s going to happen in this space. I hope not,” he said. “Unlike dot-com or crypto, most of today’s AI money sits in private hands. The public is not as exposed.”

But a correction is expected by many experts, Warren Buffett included, and will hit hardest on companies building thin wrappers on other people’s models.

“Anything that’s sort of being a wrapper on an existing model is very easy to commoditize now,” Chadhury clarified. “Companies that don’t have deep technical first principles — that’s where we’re going to see correction.”

Ram Shanmugam, a Co-Founder and Head of GTM of LayerLens, takes the contrarian angle: bubbles are good for LayerLens.

“When AI spending gets euphoric, that pushes the market towards tools that can prove what actually works. We’re not selling yet another AI platform. We’re positioning ourselves as: validate your AI investments before you make them.”

The Job Market Rewrite

The headlines scream about AI replacing jobs. Both founders push back.

“I’ve seen this cycle three times,” says Shanmugam, who lived through e-commerce, cloud, and now AI. Each wave came with the same prophecy. Each time, jobs were redefined rather than erased. “Customers are not saying, ‘I’m going to get rid of my team.’ They’re saying, ‘How do I make my team more effective using AI?'”

Chaudhury goes further. “Three years ago, to get a job at a high-tech company, you had to be very good at coding. Well, now AI can probably crack those interviews in five seconds.”

What replaces algorithmic fluency? Problem definition. Systems thinking. The skills that don’t fit on a LeetCode score. He draws the historical parallel: in the 1940s, “human computers” — women with elite mental math — were indispensable to wartime cryptography. When machines took over calculation, they became programmers. “Now you’re going to see that translate over again.”

The Plateau Question

I asked whether tech is about to get boring.

“I don’t know how you define boring,” Shanmugam says. “Boring in the tech world usually means it looks mature on the surface, but it’s super explosive underneath.” His watchlist: physical AI, AI security, sovereign infrastructure, and healthcare — an issue he cares about as someone from India who’s seen what access looks like when the system doesn’t reach you.

Chaudhury points to an Anthropic Super Bowl ad where only three out of a hundred dots — representing the American population — knew what Claude was.

“There’s a large swath of the population that hasn’t been exposed to this yet. I don’t think we’ll live through a time where tech in general for the world won’t be interesting.”

It’s a humbling thought for anyone who lives on AI Twitter. The revolution, from inside, always feels like it’s already over. From outside, it hasn’t started yet.

What LayerLens is betting on: as AI moves from flashy demo to embedded workflow, the market stops caring about novelty and starts caring about whether the thing actually works.

“Our wedge,” Shanmugam says, “is that as AI gets embedded in real workflows, the market cares less about novelty. They care more about truth, trust, and production readiness.”

In a year where every other AI pitch promises to change everything, that almost sounds boring.

Which, per the founders’ own definition, might be exactly the point.