career-ops v1.35.0, "the wider net": nine new job sources, from US school district boards to OCC Mundial in Mexico, plus a master profile you build by approving each fact from your CV.
161 PRs from 86 people, 50 of them first-timers. thank you all
https://t.co/aKJW0hNwQT
Earlier this year I stopped applying altogether. Burnt out, nothing working.
But I found out career-ops, got the job, and never left the project. User → contributor → maintainer 🫡
Mine's a small story inside @santifer's. Read his.
Companies use AI to filter candidates. I just gave candidates AI to choose companies.
I built this to find my own job: 740 listings evaluated. 68 applied. 12 interviews. 1 offer.
Six months later, I left that job to work on career-ops full time.
The talent lead of a16z's @speedrun opened a PR to career-ops this week.
He didn't ask me to integrate his talent network. He opened the PR himself: 10,000+ open roles across 200+ a16z portfolio startups, connected to the community.
Before merging, he passed the same provenance gate as anyone else: identity confirmed out of band, domain control proven cryptographically. Cleared it in a few hours without blinking.
Here's the part I keep thinking about. The engineers who run an open source, local-first job-search tool built for the agentic era are, almost by definition, the people these startups are trying to hire. The pipeline connected itself.
I didn't plan any of this. But it turns out that if you build honestly for one side of a market, the other side shows up with a pull request.
career-ops crossed 60,000 stars today.
so today the practice behind it gets a name. we wrote down how this community searches for work: six principles, and nine rights every job seeker should have. it's called the CareerOps Manifesto.
the name belongs to everyone who practices it. career-ops stays the first implementation, nothing more.
if it reads like yours, sign it. link below.
career-ops v1.19.0 is out.
~40 features, ~18 fixes since 1.18, and most of them are community PRs:
· prompt-caching + a deterministic HTML CV renderer: big cuts to token cost
· a zero-LLM JD skill-gap checker: tailor honestly, never claim a skill you don't have
· Meituan, Tencent, LaraJobs job-board providers
· Traditional Chinese + Hindi + Chinese salary-reliability tiers
still local, still free, still AI-agnostic.
https://t.co/PgEv1XNpAf
career-ops v1.14.0 is out.
18 features, 6 fixes. the big one:
openai-eval → run job evals on any OpenAI-compatible endpoint (openrouter, deepseek, glm, or local llama.cpp / lm studio). point it at a cheap model, cut your token bill.
also: first-class codex / kimi / grok support, 3 new job boards (weworkremotely, comeet, personio), and a fix for a fan-out that could burn tens of millions of tokens per run.
free. MIT. local-first.
https://t.co/IFwdIAmBmW
career-ops v1.13.0 is live 🚀
→ New boards: BambooHR + Breezy
→ Zero-token liveness checks (no more wasted tokens on dead links)
→ Dashboard upgrades + Polish locale 🇵🇱
→ Plus a web UI (2.0 RC) for the curious
Free & open source 👇
https://t.co/jCSJpkB6k4
The golden age of open source, running in a terminal near you. 👇
career-ops: free, open AI that arms the candidate. Companies use AI to filter you, we give you AI to choose them back.
⭐ 48K · MIT · yours.
https://t.co/ORxmkGzcJN
Career-ops crossed today a huge milestone, 55k ⭐️ on GitHub. Why? Hiring market is broken, and there’s need for a solution. Releasing a nice UX that make easier finding a job everyone looking for one, would be a nice next step. Check it out ⬇️
Job search tools charge $50-140/month — to people who don't have a job.
I built an AI system that evaluated 740+ listings, generated 100+ tailored CVs, and got me hired as Head of Applied AI.
Then I open-sourced it.
28K stars later, here's the master plan: 🧵