RepoJournal
Interpretability Interpretability
22 wires and counting

$ follow Interpretability

Keep up with Interpretability in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-11-07
stories 7
RSS Archive Reviews

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

param-decomp drops a public @claude agent, nnsight fixes a broken README badge

By RepoJournal · Filed · About Interpretability · Composed from the cited sources · methodology

JadenFiotto-Kaufman Michael Anderson
2 people shipped this
JadenFiotto-Kaufman @JadenFiotto-Kaufman 1 cited Michael Anderson @mganderson-goodfire 1 cited

A quiet day on the interpretability desks: goodfire-ai removed a Claude Code workflow that let anyone with write access start an agent that could push to a public repo, and nnsight repointed its DeepWiki badge at shields.io after Vercel's bot challenge broke it.

Remove the unused Claude Code workflow from this public repository goodfire-ai/param-decomp

by mganderson-goodfire

The removed `.github/workflows/claude.yaml` let any user with write access start a Claude Code agent by writing "@claude" in an issue, pull request or review, and that agent had `contents: write`, `id-token: write`, no tool restrictions, and an unpinned `anthropics/claude-code-action@beta` tag that can move to new code at any time. If you forked param-decomp, or copied that workflow into your o...

Serve the DeepWiki README badge from shields.io ndif-team/nnsight

by Jaden Fiotto-Kaufman

deepwiki.com now puts a Vercel bot challenge in front of every request, so GitHub's image proxy was getting an HTML page instead of badge.svg and rendering a broken image; the badge keeps its DeepWiki link but takes the image from shields.io.

$ ls interpretability/reviews/ # the step-back reads

the week in review nnsight 0.8.2 fixes silent wrong values under vLLM the month in review SAELens folds decoder norms at load, SAEBench JumpReLU loads fixed
Full archive → Weekly reviews →

Quick answers

What shipped in Interpretability on November 7, 2026?
A quiet day on the interpretability desks: goodfire-ai removed a Claude Code workflow that let anyone with write access start an agent that could push to a public repo, and nnsight repointed its DeepWiki badge at shields.io after Vercel's bot challenge broke it. In total, 4 commits and 3 pull requests landed.
Who contributed to Interpretability on November 7, 2026?
2 developers shipped this update, including JadenFiotto-Kaufman and Michael Anderson.
What were the notable Interpretability updates?
Remove the unused Claude Code workflow from this public repository and Serve the DeepWiki README badge from shields.io.
♥

$ log Interpretability --recent

TransformerLens 4.4.0 lands, and SAELens patches four ways its loaders misread checkpoints 19 shipped TransformerLens ships v4.4.0 3 shipped SAELens 6.55.0 runs on transformer-lens 4.x, but the pin stays 4 shipped SAELens grows a transformer-lens 4.x path but keeps the 3.x pin 4 shipped

$ log --all # elsewhere on the wire

Alignment & Post-training Anchored math grading by default in MILES, and one launcher for every dense Qwen3 TPU recipe Agent Security Sandbox Runtime 0.0.82 narrows the Windows profile read grant

$ ls showcases/

Safety Evals AI Red Teaming Agent Security Alignment & Post-training

$ init your-desk

RepoJournal can write one of these for any repo you care about. Sign in with GitHub, pick your repos and sources, and your desk gets a wire like this one every morning.