Shape Your AI, or It Will Shape You

In July 2026, Harvard Business Review published an article by three researchers from INSEAD, NYU, and Nova School of Business. Its thesis, in their words: experiment in order to shape AI rather than be shaped by it.

We've been saying a version of that sentence since before we had a company. It's printed on our philosophy: rescue the real from the virtual. AI is a second brain, not a replacement.

It's a strange feeling to watch the establishment arrive, with citations and a research budget, at a place you've been standing for a year. Not vindication exactly. More like: okay, now everyone can see it.

So let's use their map to explain what we actually build.

The HBR authors name two organizational risks from AI. Both are quiet. Neither shows up on a productivity dashboard until it's too late.

Loss of expertise. They call it decontextualization. AI can access every fact and pass the Bar exam, but it has never lived inside a problem. Data carries nuances that only a person with intuition, emotion, and real experience can read. Lean on AI to think for you, and the muscle that reads those nuances quietly atrophies.

Loss of epistemic pluralism. Language models fixate on the first idea and bias toward the average answer. Give a thousand people the same blank-slate AI and they start producing the same thoughts. HBR calls the result a monoculture. A monoculture is fragile. It misses the surprising signal — and the surprising signal is where innovation actually lives.

Two 2026 studies they cite — one from MIT Media Lab, one from Wharton — found the same thing from different angles: people using AI stop questioning its output. Cognitive surrender. You accept the wall of text because arguing with it is friction, and the interface was designed to remove friction.

Here's the part the efficiency crowd doesn't want to hear: the disease isn't "AI." The disease is generic AI.

A blank-slate model that knows nothing about you will, by design, hand you the average answer and let you rubber-stamp it. It decontextualizes because it has no context. It homogenizes because it has no you. It never pushes back because it was tuned to be frictionless.

That's not a bug you prompt your way out of. Learning to prompt better makes the average answer arrive faster. It doesn't make it yours.

Everything we make is an argument against generic AI. HBR just gave us the vocabulary for it.

Against decontextualization: Brain Kit. A persistent memory of your context — your methodology, your taste, your hard-won lessons — that your AI carries into every session. The point was never to make AI smarter. It was to keep your expertise in the loop instead of dissolving it.

Against the frictionless trap: how we onboard. HBR points to a frontier idea from another research group — "gated interfaces that unlock assistance only after the user deposits their own context." We didn't read that and build it. We built it first. Before your brain writes a word in your voice, you sit through El Espejo — a deep interview that makes you deposit who you are. Help is gated behind your own thinking, on purpose.

Against the chatbot monoculture: everything that isn't a chat box. HBR argues AI need no longer be a generic chatbot, and points to specialized reasoning interfaces. We've argued the same thing in the only language that's honest about it — design. A radar that scores signals. A companion that delivers through Telegram. Eight worlds, each with its own logic. The interface is not neutral. We treat it as a decision.

There's one idea in the article we haven't said out loud yet, and it's the sharpest one: make space where AI isn't allowed.

No-AI time blocks. Brainstorm from your own judgment before the tool is even open. Run a human team and an AI in parallel, then reconcile. One company they studied made mid-level managers hold "AI-free strategy sessions" first, so the strategy came from human judgment and AI sharpened it — not the other way around.

That's not anti-AI. That's the whole thesis in one habit. You can't shape a tool you never put down. The people who will get the most out of AI over the next decade are the ones who still know how to think without it.

The efficiency conversation is a race to the bottom. Everyone gets the same speed-up, the same average answer, faster. There's no edge in it.

The expertise conversation is where the next decade is won. Whose judgment is in the loop. Whose context the machine carries. Whether your AI makes you question more or question less.

HBR named the disease. We've spent a year building the cure. The tools are here, they're specific, and they think like you — not like everyone.

Shape your AI. Or it will quietly shape you.