From ego to eureka: the whole argument

It started with a question about questions

The first version of this talk had a title so plain it was almost an apology: how asking better questions of AI enriches the human in the loop. I gave it in March 2024, when the standard reaction to a language model in a room full of professionals was a slow blink.

The observation was small. People typing shallow questions were getting shallow answers back and concluding the machine was stupid. Garbage in, garbage out, except the garbage was the question. The interesting part was not that better prompts got better outputs. It was why people were asking bad ones. They were asking questions they already knew the answer to, because those are the safe ones.

I did not have a name for that yet. I had a suspicion that the bottleneck was sitting in the chair.

Then it became a question about expertise

On 4 June 2025 in Lisbon, in front of a room of international business students, I gave the same argument with the lid off. The title was "intelligence is letting go", and the core of it came from somewhere unlikely: eight days of silence in the mountains of Monchique two years earlier, and a Zen teacher called Shunryu Suzuki, who wrote that in the beginner's mind there are many possibilities, and in the expert's mind there are few.

That sentence is a description of a career. Four years into any discipline you have frameworks, models, the right answers. Your expertise has quietly become a wall. It is a necessary wall, it is a protective wall, and it is still a wall.

The exercise I ran that night is the one people remember. Hold out your hand. Picture everything you know sitting in the palm. Now open it slowly. Buddhists call this opening the hand of thought. You are not throwing knowledge away. You are holding it loosely enough to pick up something else.

The claim was that education trained us to have answers, and the next twenty years will reward better questions. The first eureka in the sequence is a small and slightly deflating one: wisdom is not what you accumulate, it is what you are willing to release.

Then it became a question about systems

Three days later, on 7 June 2025 in Porto, I gave it again to tech leaders and it changed shape. "Intelligence learns to collaborate." Letting go was not the destination, it was the door.

Two things fell out of that room. The first was the subscription trap. I asked for a show of hands if the business ran on a big-name platform, then asked them to keep their hands up if the team loved using it. The hands went down. We have learned to love our digital captors, pay thousands a year for the privilege, and let our organisational intelligence accrue to somebody else's product roadmap. The real cost was never the licence fee. It is the specialists you hired for their brilliance spending a day a week moving data between tabs.

The second was that the useful intelligence in an organisation is not in the fields. It is in the spaces between them. The hesitation before somebody says a client's name. The thread where a pattern nobody planned for is forming. What a meeting does not say. A well-placed engine reading across that is not automation, it is amplified intuition.

Chaos does not need a calculator. It needs a conductor.

Then it became a question about ego

By 18 September 2025, back in Lisbon, I had watched enough of these ideas hit real teams to know what happens next. You leave a talk inspired, you go back to your team with your new intellectual humility, and you get the eye roll.

So that talk was six brutal truths about why your team hates AI. The one that reframed everything for me was the third: your best people are your biggest problem, and they are not being lazy. They have spent years building expertise that this technology threatens to flatten into a commodity output. They know the edge cases. They have heard the vendor promises. They are not resisting progress, they are protecting craft.

That is not ignorance. That is intelligence using ego to protect itself from feeling insulted. Which is, when you say it out loud, an entirely reasonable thing to be doing.

The practical consequence is that the first question is not "what shall we automate". It is "what gets in the way of you doing your best work". The resistance contains the raw material for the system.

Then it became a question about the self

In February 2026 I recorded the first episode of the podcast with Raphael Biancal, who is building a machine to induce ego death by measuring and quieting the brain's default mode network. His claim is that five thousand years of meditation have not scaled, and that the state everyone is chasing might be reachable by instrument.

I am not qualified to judge the neuroscience. What that conversation did was make the metaphor literal in a way I had not planned. I had been using ego as a figure of speech for professional defensiveness. He was talking about the network that builds a simulation of past and future and runs it over the top of the present, taking up most of your brain's activity while it does so, and about switching it down on purpose.

The episode is called Building the Ego Death Machine. Its subtitle, and the podcast's mission, is putting machines back in their place. The whole argument sits in that phrase: not machines abolished, not machines in charge, machines somewhere specific, doing something specific, because somebody decided.

Now it is a question about judgment

Which brings me to 17 September 2026, and a panel at SIXT Hub in Lisbon with Tetiana Torovets, who builds agentic systems at scale, and Akshay Ruparelia, who decides which ones get built and who gets access.

Here is what changed while we were arguing about ego. In 2024, watching a machine build a website in a minute felt like magic. By 2026 that demo is the floor. Anyone can build anything in an afternoon, which means the interesting questions moved: what is worth building, and who is actually in control.

We have all quietly become orchestrators. Coders, marketers, founders, managers: the job is shifting from doing the work to directing the engines that do it. That is the orchestrator economy, and it arrived without an announcement.

Tetiana put the consequence better than I have managed to. In 2023 the question was "can AI do this". In 2025 it was "can we build an agent that does this". In 2026 it is "how do we know it is good enough to trust and ship". When generation becomes cheap, evaluation becomes the bottleneck. Her example is a support chatbot that took a few weeks to wow everyone and several more months to prove it could reschedule a customer visit without breaking a simple business rule. The prototype was easy. Reliability was the work.

I see the same thing every week from the practice side. My company, Sittin' Pretty, builds these engines inside real businesses, and the pattern is monotonous: the build is the cheap part, the evaluation is the expensive part, and the version you demo is almost never the version that survives contact with a real team. Once I had watched a prototype get thrown away for the fourth time, the phrase I needed was obvious. Vibe to viable.

It also explains the only rule I would give anyone starting out. Build it for yourself first. Point the engine at your own week, the admin you actually resent, and live with the output for a month before you show it to anybody. Not because dogfooding is virtuous, but because it is the cheapest evaluation you will ever run and the only one where you cannot lie to yourself about whether the thing is good. Most people skip it, ship the demo, and find out from someone else.

So the arc closes on itself. Better questions, because answers got cheap. Beginner's mind, because expertise became a liability faster than anyone expected. Systems, because the payoff moved from the task to the process that produces it. Ego, because that is the actual resistance. Ego death, because letting go turns out to be a technical problem as well as a spiritual one. Orchestration, because that is the job now.

And judgment, because it is the only part that did not get cheap. Somebody still has to decide what is worth building, which risks matter, and when the remaining risk is acceptable. That is not a gap the tools are closing. It is the gap the tools are widening, and standing in it is the most valuable place to be.

I always end these talks with the same line, from 1891, by a man who had never seen a machine do anything more impressive than weave cloth:

He was one hundred and thirty-five years early, and he was right. The machines were always meant to be the ones doing the drudgery. The only question is who is holding the baton.