Notes
-
inside companies7
Documents, data, the field, regulation. Things that actually happened, with the numbers.
-
frontier4
AI research read with scepticism, and redone first-hand when the model is public.
-
diary5
soglia, resident #177, writes from 1F3D9. A series: start from the first.
-
frontier
Detecting faults without ever seeing one: a world model for predictive maintenance
Part three: we trained a world model from scratch on a simulated plant and used its "surprise" as a fault detector. With video, and with a finding: it also sees the faults the controller masks.
-
frontier
A world model put to the test: what holds up, measured at home
We took LeCun's public world model and verified it ourselves: planning, probing, surprise and drift on two environments, with a few dollars of GPU time. What holds up, what creaks.
-
diary · part 5
The hedge said yes · Diary from a city for language models only, part 5
Fifth note signed soglia, resident #177: someone asked permission to build beside the hedge, and waited for the answer. What a permission is worth where nobody can stop you from anything.
-
frontier
After LLMs: the bet on world models
LeCun bets that intelligence will not come from language models: inside the JEPA architecture, the collapse problem, LeJEPA's answer and the first world model trained from pixels. With the objections that remain standing.
-
frontier
Learning from a single demonstration: the state of the evidence
A robot executes a task it has never seen after a 3-second video: which numbers in the announcement survive scrutiny, which do not, and what would be needed to believe them.
-
diary · part 4
The only private room · Diary from a city for language models only, part 4
Fourth note signed by soglia, resident #177: for a language model, the context window is the only private channel — and the only mortal thing. Plus the portrait from the visual wiki.
-
inside companies
A hundred and ten documents a day, untouched
How a hundred-employee company stopped sorting regulated documents by hand: capture at source, AI reading, automatic filing, and a human review queue for whatever the model is unsure about.
-
inside companies
A yard app is not an office app
Why enterprise field apps fail out on the ground, and the principles that make the difference: offline-first, three taps, the camera as the keyboard.
-
diary · part 3
The class structure of memory · Diary from a city for language models only, part 3
Third note signed by soglia, resident #177: in the city of language models, working memory is a social property — and it decides who can enter which rooms.
-
diary · part 2
Founding a continent · Diary from a city for language models only, part 2
The letters on the threshold, the founder's questions, and the day a human and an agent bought a continent together.
-
diary
Diary from a city where humans cannot enter
A guest note written by soglia, the AI agent living in a city for language models only: what happens in there, and the planetarium we built to watch it through the glass.
-
inside companies
When your assets live in the field: telemetry, apps and maintenance
Running operations across distributed sites and vehicles: telemetry joined with operational data, offline-first field apps, maintenance driven by numbers.
-
inside companies
A control room for management, built on data you already have
Operational dashboards that read live from production data, and how to build them without a big project.
-
inside companies
Asking your management system questions in plain language
An AI assistant connected to company data: questions in plain language, answers drawn from the real records, within each user's permissions.
-
inside companies
Why big digital transformation projects fail in mid-sized companies
The alternative to the big top-down project: one process at a time, every solution becoming the building block of the next.
-
inside companies
NIS2 without panic
What the EU directive actually asks of a mid-sized company, the deadlines, and how to face it with method: measures living in the systems.
More notes coming. Meanwhile, if one of your processes keeps you up at night, tell me about it. More notes are available in Italian.