Frontier AI PM Radar¶
A daily radar that sweeps frontier-lab blogs, arXiv, GitHub, Hugging Face, Hacker News, Reddit, product-engineering blogs, newsletters and the open web (including X and LinkedIn, via Claude's web search). It keeps the best few items, sorts them into ten categories, and turns them into a brief you can read in 20 minutes, plus a LinkedIn draft.
The aim is to compound: one brief a day, one post a day, one concept a day, and over a year that builds the judgment of a top‑1% AI product manager.
Latest: 2026-10-21¶
- (raw radar only today)
Last two weeks: 2026-10-21 · 2026-10-20
What lands every morning¶
| Section | What it is |
|---|---|
| Daily briefs | The brief: TL;DR, 8–14 stories by category (what happened / why it matters / your PM move), a deep dive with a reusable framework and an interview question, trend signals, the concept of the day, a 15‑minute exercise, and flashcards |
| LinkedIn drafts | A post draft with three hook options and a pre‑posting checklist |
| Knowledge base | A running knowledge base: every featured item filed by category, newest first |
| Concept glossary | Glossary of the concepts of the day |
| Flashcards (CSV, on the knowledge page) | Every flashcard, ready to import into Anki (separator ;) |
| Weekly reviews | Sunday review: what mattered, trend lines, a weekend build, and a long‑form post |
| Structured data (in the repo) | The structured digest, for search or your own tools |
Categories: Agents & Agentic Systems · Evals & Quality · Governance, Safety & Policy · Models & Capabilities · AI Product Design & UX · Infra, Cost & Latency · Research Frontier · Open Source & Dev Tools · Industry Moves & Strategy · PM Craft & Playbooks
How it works¶
collect (55+ feeds, page watchers, arXiv, HN, GitHub, HF, Reddit, optional X/LinkedIn via Apify)
→ dedupe against 90 days of history
→ pre-rank (source weight × popularity × PM-topic keywords, capped per source)
→ triage ~140 items with Claude Haiku 5.5 (0–10 score + category)
→ web scout: Claude Opus 5.5 + web search finds what the feeds missed
→ synthesize the top ~40 into the brief with Claude Opus 5.5 (structured output)
→ write markdown, update knowledge base + README, commit
It runs on GitHub Actions at 10:30 UTC every day (.github/workflows/daily.yml), and on Sundays it also writes the weekly review. If the Claude step fails, a raw ranked list is published so no day is lost.
Train with it¶
- The plan: daily and weekly routine, LinkedIn strategy, roadmap
- The curriculum: 10 competencies of a top-1% AI PM, with a hands-on ladder
- Source code and sources list