Observed arrival · 2026-10-15
Inside the Model’s six-question transformer lab
A browser-based set of guided experiments for investigating how a small transformer produces predictions.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Developers learning transformer internals
- Worth noticing
- The browser engine is checked against an independent NumPy implementation to within 1e-9; backpropagation is checked with finite differences.
Field notes
The investigations progress from identifying a component involved in a prediction to designing controls and assessing causal claims. The page says learners can trace a token through 26 operations, moving between calculated values, equations, and reference code. It also offers downloadable weights for rebuilding the forward pass in NumPy, while cautioning that the teaching model is not representative of a production LLM’s capabilities.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
ƒJavaScriptBrowser-side code central
One card from the complete issue