Observed arrival · 2026-09-30
Token Trails: language models, one token at a time
An interactive guide follows how language models process text, covering their architecture, training, serving, and agent workflows.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Readers learning language-model internals
- Worth noticing
- Its topic map names specific methods and models, from Mamba and Mixtral to paged attention and speculative decoding.
Field notes
The topic map runs from foundations such as tokenizers and embeddings through post-training methods including SFT, RLHF, DPO, and LoRA. It also groups agent patterns such as ReAct, tool calling, RAG, and multi-agent handoff, giving readers several specific routes through the material rather than one linear overview.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
One card from the complete issue