TechnicalAI glossary
Chunking
Chunking documents before indexing them for a RAG. It's not a detail: this is where most of the quality is decided. Too small, chunks lose their context; too large, they bring back noise.
In other words
Splitting a book into index cards. If you cut mid-sentence, none of the cards make any sense anymore.
Alyriaz articles that mention it
- AI: 75,000 fewer permanent jobs a year for young people
According to the Banque de France and HEC, permanent-contract hires among 21-29 year-olds have fallen by 28% in roles exposed to AI. Newcomers to the job market are paying the price first.
- Claude Sonnet 5.5: nearly Opus, at half the price
Across 44 professions, Claude Sonnet 5.5 nearly matches Opus 5.5 at half the price. What it changes for you, and where the top model keeps its edge.
- Why ChatGPT is free when it costs a fortune
An AI response costs electricity and hardware. So why are so many services free? We explain who pays, why, and how long it can last.
- Claude Opus 5.5: flagship-level performance for 60% less
Anthropic has released Claude Opus 5.5, claimed to be on a par with Fable 5.1 at a far lower cost. We analyse the figures, the comparison with GPT-6 Astra, and the tension with the call for a slowdown.
- Six hundred billion parameters at a dollar per million tokens
StepFun has launched Step 5 Preview, a 600-billion-parameter model with a one-million-token context window, priced at 1 dollar for input. We put that rate into context within the current landscape.
- Suno, Udio and the Sacem: who gets what on an AI-generated track?
Generating a track with Suno or Udio does not make you an author under French law. We explain the difference between a commercial licence and copyright, what you can actually declare to the Sacem, and why the Munich court ruling of 31 July could reshuffle the deck.