Cédric · AI technology
I take AI technology apart until I understand it.
Currently surveying every model on the market.
01 — Knowledge
Knowledge is
the part that
stays.
Models arrive monthly and are superseded within a year. Understanding does not expire on that schedule.
So this is a working notebook rather than a brochure. What I learn about AI technology, written down in my own words, kept somewhere I can come back to it.
Not a summary of what other people said. My own model of how a thing works — including the parts where that model is still wrong.
The point is not to be current.The point is to understand well enough that being current stops being the hard part.
02 — Now
Surveying every model on the market.
The current work is a survey: what exists, what it claims, and how the pieces relate to each other.
A projection, not a measurement. Every point here is generated. The cloud lives in four dimensions and you are looking at a two-dimensional slice of it — drag to rotate the projection plane and watch the structure reorganise.
03 — Commitment
I take on the projects worth the trouble.
Special projects get special attention. If an idea is interesting enough, I go deep on it rather than wide.
- a
Deep over wide
One problem, properly understood, beats ten that were skimmed.
- b
Written down
If it is not noted, it did not happen. Notes are how the work survives.
- c
Built to last
Tools that outlive the demo they were first written for.
04 — Direction
Further into AI.
Not one model. Not one framework. The field.
Where it goes next is what I want to be close to. This is where I keep what I learn on the way.