Sushant's Open Notebook
All Tracks
Follow by RSS →Track ICore LLM
↗What enters a language model, how tokenization shapes it, and where the costs begin.
- 01An LLM has never seen a letter in its life
- 02The hidden language tax
2 written0/2
Track IIAgent Patterns
↗Observable steps, narrow tools, retrieval, and clear recovery paths.
- 01Observable steps
- 02Narrow tools
- 03Retrieval inside an agent
- +1 more
Planned
Track IIIInference
↗One request through batching, memory, latency, and the economics of serving it.
- 01One request, end to end
- 02Batching
- 03Memory
- +2 more
Planned
Track IVMachine Learning
↗The work between a training set, a reliable evaluation, and a model worth using.
- 01The training set
- 02A reliable evaluation
- 03A model worth using
Planned
Track VData Analysis
↗A precise question, the evidence for it, and what the data cannot support.
- 01A precise question
- 02The evidence for it
- 03What the data cannot support
Planned