Dev's Multica 1.0 build log: 432M tokens, 89 agent runs, just $183

jiayuan_jy · x · 2026-10-31

Developer jiayuan shared a dashboard of building Multica 1.0 with coding agents: 89 runs over 2.5 days, 432.3M tokens consumed (98% from cache), 13h44m of agent time, for a total of just $183.

By agent, Lambda did 98% of the work ($180) vs Emacs ($2.84). By model, claude-opus-5.5 high accounted for $152 and the xhigh tier $28.23, with GPT-6.1 Sol at $2.84. A run-level breakdown shows a 98% cache hit rate saved $47.39 on a single run — a vivid look at how prompt caching cuts long-horizon agent costs.

Original post →

More from coding & agent

coding & agent channel →