Open-source Jive rebuilds the agentic loop around "System One" graph calls, beating Codex on speed and cost 7x
burakyildizdev · reddit · 2026-09-25
A developer open-sourced Jive, a harness built around a redesigned agentic loop: let the LLM's full power handle planning, then execute the plan with cheap "System One" fast calls instead of burning a full LLM call on every step.
Core mechanism: graph calls replace tool calls—each graph is a DAG of bash nodes and jev nodes with dependencies and output references; the agent maps its own execution flow while reasoning, then runs it via jev calls.
Benchmarks (mean, Jive vs Codex vs Claude Code): 2m31s/8.7k tokens vs 17m40s/16.2k vs 12m12s/32.7k. Jive wins all six tested tasks, e.g. conversationeval at 3m26s/11.1k vs Claude Code's 16m48s/47.9k.
Good at: repo investigation, bulk classification, multi-step profiling, repetitive edits, eval workflows. Author notes accuracy isn't yet verified on large-scale SWE benchmarks. Repo: github.com/merijjeyn/jive.
Related event: Open-source Jive Rebuilds Agent Loop with Fast/Slow Thinking(2 posts)→
More from coding & agent
- Why inference speed will be AI products' biggest competitive edge in 1–2 years — itsOmSarraf_ · 2026-11-11
- What happened when I moved vector retrieval into Snowflake SQL instead of orchestrating RAG outside — dima806_dima · 2026-11-11
- OpenAI engineer runs Codex for 26 days straight hunting Atlantis, spawning "vibe archaeology" — Polymarket · 2026-11-11
- He replaced a $5,000/month ghostwriter with 11 Claude Skills and earned $47,000 from one post — aryanXmahajan · 2026-11-11
- Refero Styles ships 2,000+ DESIGN.md files to give Claude Code real design taste — bbssppllvv · 2026-11-11
- Agent circuit breaker: openclaw shorts bot halts itself after 3 failed experiments — Rich_Standard_1724 · 2026-11-11