BUILD · RESEARCH INTELLIGENCE · 2026
Topic Intelligence Agent
Does the research and separates the facts from the hype.
Give it a topic and it pulls from Hacker News, Google News, Reddit, arXiv, and similar sources, then sorts what it finds into what's actually established versus what's just confidently repeated opinion — and flags what nobody's talking about yet. It also has skills to help shape that research into a point of view, and capture it in a form I can share.
Internal tool - no public link
QUICK FACTS
Role — Solo. Design, build, ship.
Stack — Claude Code agent, three custom skills (topic-scan, pov-builder, content-creator). ources include Hacker News (Algolia API), Google News RSS, Reddit RSS, and arXiv's API, chosen per topic.
Status — Internal tool. In active use.
Focus — Doing real research on a topic and separating fact from hype, fast.
THE SHIFT
Before it feels like: everything online reads like consensus, and it's hard to tell what's actually established versus just repeated confidently enough times.
After: I know what's solid, what's still speculation, and where the real gaps in the conversation are — before I form an opinion, not after.
How it works
1. Give it a topic — a keyword, an article, a URL, a transcript, or a question.
2. topic-scan queries Hacker News, Google News, Reddit, and arXiv (source mix depends on the topic) and maps what's actually being said.
3. It sorts claims into Solid (evidence, broad agreement) and Opinion/speculation (asserted but not established).
4. It flags the angles that are under-discussed — the real gaps, not just the loudest takes.
5. From there I can build a POV and a draft, but the research is the part doing the real work.
Why I built it
Most "research" I did manually turned into a pile of takes that all sounded confident, with no real sense of what was established versus just repeated. I built this to do that separation properly — sort fact from hype, and flag what the conversation is missing — so I'm not mistaking consensus for evidence.