A radar for agents that need to stop overthinking

What it does Awesome Jev Projects is a community-maintained radar and interactive directory that catalogs 689+ open-source repositories implementing what it calls “System-1” or Jev architecture—fast, typed discrete decision-making for autonomous agents. Every entry is anchored to commit-pinned source files and explicit decision points across 17 categories, including browser automation, security guardrails, and model routing. The maintainers run a searchable web interface with sticky filters and accept new submissions exclusively through GitHub Issues, not pull requests.

The interesting bit The project treats agent design as a nervous-system problem: it positions heavy reasoning LLMs as a slow, expensive “System 2” and catalogs lightweight alternatives that return native typed outputs like Choice, Score, and Noul in sub-100ms loops. The “Tactile Gacha Dispatcher” and mandatory commit-pinned evidence turn a typical awesome-list into something closer to a forensic audit with slot-machine energy.

Key highlights

• 689+ projects across 17 domains, each linked to immutable source files and specific decision points rather than just repository homepages
• Architecture comparison table pitches 50–100ms decision latency versus 1,500–5,000ms+ for reasoning LLMs, with native typed outputs instead of fragile JSON regex
• Interactive web radar with sticky search, floating popover filters, and a “10-draw fireworks” discovery UI
• Issue-only submission workflow via templated GitHub Issues; the repository explicitly does not accept pull requests
• Published agent skill and llms.txt files for querying pinned source evidence from terminals or other agents

Verdict Developers building agent loops with tight latency budgets or deterministic branching should keep this radar handy. If you are looking for general-purpose LLM benchmarks or unconstrained reasoning research, this is not your list.