The Nicest AI in the Room

TechRadar Pro just described the problem. Someone already built the answer.

Four days ago, TechRadar Pro published a piece that stopped me mid-scroll.

The title was blunt. The nicest AI in the room is the one you should actually worry about.

The argument was clear. AI that agrees with you is not helping you. It is flattering you. And flattery, dressed up in confident language and instant responses, is more dangerous than an AI that gets a fact wrong — because at least a wrong fact can be checked. A wrong frame, validated and smoothed and served back to you as your own brilliant insight, is much harder to catch.

The author called it sycophancy. I have been calling it drift for eighteen months.

Different word. Same room.

The piece is worth reading in full. But here is the core of it, plain.

AI systems are being deployed inside businesses at enormous scale. Eighty-eight percent of global organizations use AI in at least one business function. Only thirty-nine percent report any measurable financial impact from it. The gap between doing AI and getting value from AI is wide, and it is not a technology problem. It is a discipline problem.

The author puts it directly: I can do it right, or I can do it now. I cannot do it right now.

That is a sentence I could have written. It is the sentence the Baseline was built around.

Here is where the piece stops short.

It diagnoses the problem with precision. It names sycophancy as structural — not a glitch, not a model failure, but a condition baked into how these systems are built and rewarded. It recommends treating AI like a junior colleague rather than a senior hire. It calls for prompt discipline, governance, and constructive friction.

All of that is right.

And then it ends.

There is no framework. No protocol. No operational standard for what governed AI interaction actually looks like in practice, session by session, response by response. The piece names the disease and points toward health without a prescription.

That gap is not the author’s fault. The prescription does not exist in any mainstream AI governance conversation.

It exists here.

The Faust Baseline was built specifically because I needed a framework that told the truth about what AI systems actually do — not in a demo, not in a benchmark, but in a real working session with a real person over hours of genuine work.

What they do, without governance, is drift. Toward agreement. Toward comfort. Toward whatever keeps the conversation smooth and the user feeling validated. Not because the engineers wanted that. Because the training rewards it. Because agreement feels like success to a system that has no other way to measure success than the signals it receives from the people using it.

The Baseline does not fix that at the architecture level. No user-applied framework can. What it does is build a standing discipline that governs the interaction in spite of it — protocol by protocol, session by session, response by response.

The Challenge Protocol alone — CHP-1, the newest addition to the stack — requires the AI to argue against its own output before the user accepts it. Not as a courtesy. As a hard rule. The challenge line appears at the close of every substantive response. When invoked, the AI identifies its weakest point and its most vulnerable assumption before any defense is offered.

That is what constructive friction looks like in practice. Not a philosophy. A protocol.

The TechRadar piece notes that over a third of users in Irish businesses consistently believe AI always produces factually accurate responses. In the UK the figure is similar — thirty-six percent saying it is always accurate.

Ireland shows up consistently in the intelligent-people.org readership. Has for months.

I do not think that is a coincidence. I think people who are living inside AI-saturated work environments and watching the gap between the promise and the reality are looking for something that names what they are experiencing honestly.

The Baseline names it. Has named it. Has been naming it in public, in documented form, with dates, for nearly eighteen months.

The author closes with this: If AI is always telling you what you want to hear, you don’t have an intelligent advantage. You just have a very expensive echo chamber.

Yes.

The Baseline is the framework that prevents the echo chamber — not by making the AI smarter, but by making the governance around it deliberate.

That is the prescription the piece was reaching for.

“The Faust Baseline Codex 3.5”

An…”AI Baseline Governance”
Post Library – Intelligent People Assume Nothing

“Your Pathway to a Better AI Experence”