Day 17 - Casino del Masnou: Palantir, Harari, and bureaucracy that talks

Notes from the Casino del Masnou meetup: why Palantir worries us, what it does well with data, and how AI already outwrites humans and finds gaps in procedures.

On 10 July 2026 we met at the Casino del Masnou (Carrer de Barcelona, 1, 08320 El Masnou) to continue the conversation announced in day 16: Palantir as a real case, and AI + bureaucracy (Harari again), this time with language-psychology nuance.

It was not a product pitch. It was an informal table: concrete risks, respect for what that company does well, and a shared sense that models already draft and “read” procedures better than many humans.

Meetup facts

Date: 10 July 2026 (follow-up to the day 16 announcement).
Place: Casino del Masnou, Carrer de Barcelona, 1, 08320 El Masnou, Barcelona. Map.
Venue site: casinomasnou.com.
Axes: Palantir (data infiltration / mapping) + AI that outwrites humans and “hacks” bureaucratic gaps.
Venue note: day 16 announced Bar Pekin; the meetup took place at the Casino.

Palantir: danger and competence

We talked about Palantir as a pattern, not a meme:

Wide infiltration: the business model pushes into as many organisations as possible (public and private). The bigger the graph, the more useful it becomes.
Data on thousands (and millions) of people: you do not need phone spyware to worry; integrating systems that already hold identity, filings, relationships, and movement is enough.
Database mapping: what it does very well is joining silos, modelling entities and relationships, and making a map visible that nobody held in their head before. That is operational power; it is also concentrated power.

The table question was not “is it legal?” but “who controls the map when a third party owns the map?”. For KM0 the contrast is local: Cloud and mail in the EU, public price, no selling the profile as the product.

Harari again: AI already writes better

We returned to Harari because day 14 left the thesis open: civilisation runs on bureaucracy (law, forms, registries), and AI can operate inside that language at scale.

At the Casino the nuance was more psychological:

Models understand (or simulate understanding) words more consistently than a tired human: tone, templates, exceptions, gaps.
Writing “better than us” is not only style: it is closing procedures, drafting reports, filling holes, and proposing the next step.
“Hacking” bureaucracy here does not mean cracking a server: it means finding cracks in the text (ambiguity, exceptions, deadlines, poorly defined forms) and using them with precision.

If bureaucracy is a word game, whoever masters the language of the game has the edge. That is frightening and, at the same time, explains why tools like Palantir gain ground: they do not only store; they interpret and connect.

What we take back to KM0

Critical watch on platforms that map whole organisations.
Technical respect: good data mapping is hard; slogans do not defeat it.
Practical preference: nearby, auditable infrastructure with a public price (pricing, KM0 Cloud).
Keep talking in person: meetups.

Close

This day makes clear where we met, what was said about Palantir (risk + competence), and what was said about AI/Harari (writing + bureaucratic gaps).

Next: day 18 covers a professional session on KM0 brand and style (and why the brand then had no soul).