Whoever asks about my work — a person, a search engine, an agent — reaches the same answer, because every surface they can reach derives from one model I keep, and no claim in it stands without its evidence.
Generated from the model, so the words below are its own — and in the one language it is written in. The rest of this site is bilingual; a translated copy would be a second thing to keep true, which is what this page argues against.
Everything public about my work is derived rather than maintained. One model holds the facts: what I have built, which capabilities I claim and on what evidence, and what I hold to. The CV and its per-application dossiers, the website and its talks, the LinkedIn profile and the loadable skill bundle are outputs of it. When two of them disagree the model is what gets corrected and the surfaces are rebuilt; none of them is edited on its own.
The claim is not that any single surface is good. It is that they agree — because someone asking about me increasingly asks a machine, and a machine assembling an answer from contradictory sources produces either mush or a confident error. I would know this holds when the same question put to a person, to a search engine and to an agent comes back the same.
The scope is the work. Private life is out of scope by design rather than by omission: the schemas have nowhere to put it, which is a stronger guarantee than a promise.
Naming what is broken is free, so it is not a position until something works differently.
I do not publish a criticism until I have built the thing that answers it, and then the working alternative is what I lead with — the argument comes second and can be lost without the point collapsing. Where I cannot build it, I say the problem is real and leave the criticism unmade.
I never let a complaint stand as a position.
AI moved the constraint from building quickly to deciding correctly; the decision is the work.
I write the decision down before the code — with the alternatives that lost and why — so the next person can disagree with a reason rather than a rewrite. I spend the time that fast building saves on choosing well, because the constraint moved there.
I never ship a feature in a day and argue about it for a quarter because nobody can say what it was for.
A platform that outgrows the team that runs it is a liability with good uptime.
I grow the organization as deliberately as the platform — hiring, practices, ownership — so that every part of the system has someone who could be woken up for it and would know what to do. I judge a platform by the team that can run it without me.
I never let five people know everything while twenty wait for them.
A business described as concepts and capabilities is one that people and AI can both act inside of.
I keep a written model — domains, capabilities, processes, rules — that the code, the org chart and the agents all refer back to, and I keep it where the people who own the facts can edit it. I build from the model, not beside it.
I never let the model live in three heads and a slide deck, with every system encoding a slightly different version of it.
A platform counts when it runs for customers under real security and compliance obligations, not when the demo works.
I own reliability, incident management and the audit trail as part of building the thing. I treat ISO 27001 and GDPR as design inputs, not a form to fill in after. I call it done when it runs for customers under real obligations, not when the demo works.
I never meet a launch date with a system nobody is on call for.
Generated from robertblust/mental-model@135d7b4 — model/vision.md and model/values/. That repository is an instance of CompanyGraph, a meta-model for describing a company as a graph of Markdown.