The call-your-bluff culture in science and engineering

There is one thing I genuinely love about science and engineering, and it is not the technology. It is that somebody will call your bluff.

Not politely. Not in a follow-up email. Right there, in front of everyone, while you are still mid-sentence.

The compiler does not care who you are

This is the part outsiders never quite believe. In an engineering room, a graduate two months into the job can put a stack trace on the screen and end an argument with a chief technology officer. Not because the graduate is braver, but because the machine already settled it. The build passed or it did not. The query returned in 40 milliseconds or it returned in 12 seconds. There is no seniority in a benchmark.

Science runs on the same cruelty, just slower. You publish, and somebody on another continent who has never met you tries to reproduce your result and cannot. Your title does not travel with the paper. Your reputation buys you a few months, maybe, and then the experiment answers for you.

I have been on the receiving end of this more times than I would like to count, and every single time it made the work better. That is the trade. You give up the comfort of being unquestioned, and you get to actually be right.

Now walk into the room where the money is decided

Different physics entirely. Somebody presents a slide claiming a 40% efficiency gain and nobody asks 40% of what, measured how, against which baseline, over what period. The number is not evidence, it is decoration. Everyone nods. The deck gets forwarded. Six months later a different consultant presents a slide explaining why the first project underdelivered, and nobody calls that bluff either.

We have built an entire professional layer where claims are never tested, only repeated. Awards judged by the people who paid to enter. Panels where four speakers agree with each other for forty minutes. LinkedIn posts announcing transformations that are, in the building, a scanned PDF in an automated email.

The difference is not intelligence. The people in that room are frequently smarter than the people in the engineering room. The difference is that nothing in their environment is designed to say no to them.

AI just made the bluff cheaper

Here is what worries me now. Generating a confident, well-structured, professionally-worded claim used to cost something. You had to at least write it yourself, which meant passing through your own doubt on the way. That cost is now approximately zero.

A model will happily produce a fluent architecture review, a convincing risk assessment, a benchmark table, an executive summary of a system it has never run. It will do it in eight seconds and it will look better than what your actual engineer would have written, because your actual engineer hedges where the truth is uncertain, and the model does not.

So the volume of unverified claims is going up sharply while the number of people willing to call the bluff stays flat. You do the arithmetic on where that ends.

What calling the bluff actually looks like

It is not aggression and it is not cynicism. It is four or five boring questions, asked out loud, every time:

• What was the baseline, and who measured it?
• Can I see it fail? Show me the case where this does not work.
• What did you evaluate this on, and did you write the evals before or after you saw the results?
• If this is running in production, what broke in the first month?
• Who disagreed with you internally, and what did they say?

That last one is the sharpest. Any real project has an internal dissenter. If somebody tells you nobody disagreed, either they are not telling you the truth or they built something so small that nobody bothered to have an opinion about it.

The regional version of this problem

In our part of the world this is harder, and I am not going to pretend otherwise. Calling a bluff carries a social cost here that it does not carry in a Berlin engineering standup. Hierarchy is real, seniority is respected, and publicly telling a senior person that their number is not a number is expensive.

I understand the instinct. I just think we are paying for it. Every unchallenged claim compounds, and eventually it becomes a system nobody wants to touch, running on assumptions nobody has tested, defended by people whose credibility is now attached to it being true.

The fix is not rudeness. It is making the question normal. If the first thing anyone asks about any claim is “what is the baseline”, then asking it stops being an accusation and starts being the weather.

Why this matters more than it sounds

Engineering and science are not more truthful than other fields because engineers and scientists are more honest people. They are not. They are more truthful because the culture makes lying expensive and being wrong survivable. Those two things have to come together. If being wrong ends your career, everyone hides, and then the bluffing starts.

That is the culture worth protecting, and it is worth exporting into every room where a decision costs real money. Especially now, when a machine will write you a beautiful, fluent, entirely unverified paragraph about anything you want.

So when was the last time somebody called your bluff and you thanked them for it? And when was the last time you called someone else’s?