← All tools

AI Radar

A technology radar that keeps itself current.

The Thoughtworks radar is a strong mental model with a poor refresh rate. Twice a year is fine for JVM frameworks. For AI tooling the map is stale before the ink dries. AI Radar closes that gap by computing the map instead of curating it.

When a landscape moves faster than your review cadence, curation turns into Cognitive Debt: the organisation stops forming its own judgment and starts deciding by noise. AI Radar keeps the judgment in-house and the map current.
AI Radar's live scope view: tool contacts plotted by category bearing and momentum, with emergent clusters listed in the sidebar
The live scope. Every contact is a repository, placed by momentum and category bearing. The clusters in the sidebar, Harness, MCP server, Coding agent, were not defined by hand: they emerged from the embeddings.

Trending GitHub repositories are re-scanned every 30 minutes and streamed to open browsers over SSE. Categories are not hand-assigned: they emerge from the data through an embed, reduce, HDBSCAN, c-TF-IDF pipeline, and the Clusters view shows the exact steps that produced them. Adoption rings (Adopt, Trial, Assess, Hold) are computed from momentum and maturity, not voted on in a committee.

Self-hosted in one docker compose up. The radar belongs to your organisation, not to a vendor. The observation scope is configurable: AI and dev tools out of the box, Rust or platform and DevOps as bundled presets, or your own areas added through the UI with no rebuild.

A curated vendor PDF gives you a map you cannot audit, cannot extend, and cannot keep current. AI Radar is nachvollziehbar (the pipeline is open), unabhängig (self-hosted, MIT), and current (recomputed every 30 minutes). Those are three properties a hand-curated landscape cannot offer.

Python 3.12FastAPIPostgres + pgvectorUMAPHDBSCANNext.js 16Docker
Live demo ↗ Code (MIT) ↗

What is AI Radar?

AI Radar is an open-source, self-hosted technology radar that keeps itself current. It re-scans trending GitHub repositories every 30 minutes, clusters them by embedding, and plots them on adoption rings computed from momentum and maturity.

How is it different from a Thoughtworks-style radar?

A classic radar is curated by a committee and published twice a year. AI Radar is computed from data and recomputed every 30 minutes, so the map never lags the field it describes.

Can I use it for domains other than AI?

Yes. AI and dev tools ship as the default. Rust and platform/DevOps are bundled presets, and you can add your own observation areas through the UI without a rebuild.