← Home / 2026-Q4
Period

2026-Q4

10 essays
Most recent
2026-11-14

Web Summit: Have Tech Trade Shows Become Instruments of Soft Power?

Learn to read a tech trade show beyond its stage: who organizes it, who funds it, why it takes place in that city. Drawing on the opening of Web Summit 2026 in Lisbon, and on Portugal, Qatar, Brazil and China, separate the showcase, the network and the editorial decision.

2026-11-07

Malorean Corp. is now a book

When anyone can do what used to belong to another profession, who answers for what? The book that just came out answers with the Guarantor: for each domain, someone answers for its coherence without validating everything or doing everything. You come away with a way to keep autonomy without letting it dissolve into chaos or drown in approvals.

2026-10-29

Ontology-Driven Design: what if the ontology became the safety net for AI-generated code?

When an AI agent writes the code, the tests pass but the business logic can drift without anyone noticing: two names for one concept, a rule contradicted, an abandoned notion creeping back. An ontology extracted from the code, held up against the intent before each change and against the code afterwards, makes that drift visible and tells an intended change from a mistake.

2026-10-28

Smoothing over contradictory customer feedback is a product mistake

A summary like "simplify while making the data reliable" gets everyone to agree and makes the real product problem disappear. Keeping each piece of feedback intact, one idea per card, and writing the tension down as an idea in its own right turns an awkward contradiction into a design question — and turns an AI, which smooths things over by default, into a guardrail against smoothing.

2026-10-28

Start from scratch or copy: two ways to lose what you know

Answering the same security questionnaire for the tenth time without redoing everything, and without copying a wrong answer. Filing facts by topic, with their origin, their date and their limits, lets a 115-question questionnaire be 80-90% prepared in ten minutes, and turns whatever is missing into a precise question sent to the right team.

2026-10-28

Nobody knows how to pick a case back up

Twenty minutes at every meeting just getting everyone back on the same page, a handover that falls apart when someone leaves: you don't pick a shared case back up by rereading your inbox. A pickup page — where things stand, what's fragile, a single next action — kept up while the work happens lets anyone, an AI included, start from the same point in a minute.

2026-10-28

A deliverable keeps the conclusion and loses the why

A recommendation that worked for one client can be wrong for the next, and the report won't tell you why. Keeping each decision with its alternatives, its reason and its conditions, then drawing notes from it with no client data, lets a firm reuse what it has learned without betraying anyone — and see on day one when a good answer no longer applies.

2026-10-14

The `Context` is now a book

Your conversations with an AI keep the conclusions and lose the reasons: the option you ruled out, the objection left open, why you decided the way you did. The book that has just come out describes the workspace built against that — one per subject, where the AI can trace an idea back to its origin and pick the work up where it stopped.

2026-10-20

Graphify without AI: turning a wiki into a navigable graph for zero tokens

A thousand linked notes and no overall view: no way to see which ideas are central, which are isolated, or how two distant concepts meet. When every link in a wiki is already explained inside the notes, a 160-line Python script is enough to build a complete graph, explored with Graphify, in 0.2 seconds and without spending a single token.

2026-10-04

Stateful Corpus Research Loop: stop collecting sources

Knowing what to look for next, and above all when to stop. Instead of fixing 30 or 50 sources up front, you start from a small vetted corpus, turn it into structured knowledge, and let what it knows — and what it doesn't — pick the next search. The result: less redundancy, visible blind spots and a stopping criterion you can test.

Most recent articles