neuron-explorer
Qwen3.5‑4B 전용 조향형 뉴런 탐색기, 모든 MLP 뉴런을 밀어 결과를 보관·탐색
아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
프로젝트 폴더에서 /nomorevibe verify를 실행하면 소유권을 확인하고 소개를 직접 관리할 수 있습니다.
1분 만에 확인하기README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
A k9s-style terminal UI that runs Claude Code agents against your issue tracker. Pick issues, dispatch an agent per issue into its own git worktree, and watch a live board of what they're doing — questions answered inline, draft PRs, CI, review state, cost.
It runs on Claude subscription auth (the logged-in claude CLI), works on Linear today behind a provider interface, and never touches your working copy: agents get worktrees, and they only ever open draft PRs.
📖 Read the docs · 🚀 Getting startedREADME 전문 보기 ↗A k9s-style terminal UI that runs Claude Code agents against your issue tracker. Pick issues, dispatch an agent per issue into its own git worktree, and watch a live board of what they're doing — questions answered inline, draft PRs, CI, review state, cost.
It runs on Claude subscription auth (the logged-in claude CLI), works on Linear today behind a provider interface, and never touches your working copy: agents get worktrees, and they only ever open draft PRs.
📖 Read the docs · 🚀 Getting started
Disclaimer: This is an app I find productive and have developed almost blindly with claude. I've done typescript development, but this is far out of my expertise and the code could very well be vibe-coded trash. I don't really care. The goal is to get something out that is productive. I'm also adding features and fixes to it regularly.. while I am putting thought into preventing backwards compatible issues I have no guarantees. Use at your own risk, at this point in time I'm not accepting contributions from the general public, but feel free to open an issue.
Every task is also a searchable, sortable table — same data, same keys, :tasks:
Both frames are coli demo — a fabricated board you can run yourself in thirty seconds.
git clone https://github.com/jubrad/colinear && cd colinear
npm install && npm link
coli demo # a populated board — no account, no agents, nothing billed
git clone https://github.com/jubrad/colinear && cd colinear
npm install && npm link # builds dist, puts `coli` on PATH
coli init # pick `sqlite` to try it with no account, or `linear` for the real thing
npm run doctor # claude CLI, gh auth, key, repos
coli # `?` for help, `:` to jump between views
Requirements: the claude CLI logged in (leave ANTHROPIC_API_KEY unset) and gh authenticated. A Linear API key too, unless you start with the built-in local tracker.
:todo pulls reviews, waiting agents, your issues, milestones and projects into one list, ranked by an agent against a policy you can rewrite: reviews within a working day, deadlines, then priority.jubrad.github.io/colinear — every page below, with search and navigation.
| Getting started | install, configure, first dispatch |
| Security & blast radius | what it can touch, what it can't, where your data goes |
| Demo mode | coli demo — see it working before configuring anything |
| How dispatch works | triage → work → checks, messaging agents, sessions |
| Configuration | every option, with defaults |
| Remote daemon · Docker | run the daemon on a VM or in a container (work in progress) |
| CLI | coli, init, daemon, gc, contexts |
| Views | one page per view — :board, :issues, :reviews, … |
| Architecture | how it works inside, and the gotchas |
Used daily by its author against a real Linear workspace; APIs and keybindings still move. Two issue providers today: Linear, and a local sqlite tracker for trying it without an account — the interface and its capability flags are what a third one plugs into.
MIT licensed.