ddosi
Featured
PacketLens
PacketLens is a pure front-end, offline pcap analysis tool that runs entirely in the browser, supporting deep protocol decoding, HTTPS decryption, and million-packet instant loading without any backend.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It brings Anthropic's research on the J-space (global workspace) to an interactive, continuous chat interface, allowing anyone with a GPU to watch the model form judgments, plan ahead, and hold intermediate reasoning tokens before they appear in output.
Who it is for
- AI interpretability researchers
- ML engineers curious about internal model reasoning
- Enthusiasts and hobbyists exploring LLM transparency
- Educators demonstrating transformer inner workings
Use cases
- Observing when a model forms a verdict during reading, before any reply tokens appear
- Studying planning behavior where the model holds high-strength concepts while emitting unrelated tokens
- Debugging model reasoning by replaying sessions and inspecting per-token, per-layer readouts
- Creating reproducible, GPU-free demonstrations of internal workspace dynamics for presentations
Strengths
- Conversational and continuous: renders lens readouts during both reading and generation phases
- Streams at generation speed via KV cache and GPU, with low overhead
- Sessions can be exported as JSON and replayed in any browser without a GPU
- Validated against the reference JacobianLens implementation with top-5 readout matches and logit cosine similarity ≥ 0.99998
Considerations
- Only single-token concepts are visible; multi-token concepts appear fragmentary or invisible
- The lens captures only the approximate J-space workspace, not the full internal state, and layers outside the fitted range are not observed
- Requires an NVIDIA GPU with ~10 GB VRAM or Apple Silicon with 16+ GB unified memory; CPU fallback is very slow
README quick start
Getting started & staying tuned with us.
Related repositories
Similar projects matched by category, topics, and programming language.
ViffyGwaanl
Featured
kimi-k3-learn
An interactive learning system that turns the 47-page Kimi K3 technical report into a single offline HTML file with running algorithms, 3D visualizations, spaced repetition quizzes, and a smart highlighting QA tool.
iamtechartist
Featured
human-cell-visualizer
An interactive 3D visualization of three human cell types using Three.js, WebGL particles, and custom GLSL shaders, allowing users to rotate, zoom, morph, and explore annotated structures.