Back to Claude Code After 6 Months: Opus 5.5 Sips Quota, ultracode Cracks Memory Leaks
Hesamation · x · 2026-10-11
Hesamation returns to Claude Code after 6 months: Opus 5.5 is so efficient that a week of long tasks on a Max 20x plan burned only 50% of the quota; ultracode found a memory leak root cause that Astra couldn't (though it's painful for implementation work); agent-to-agent messaging is a hit, and the desktop app now beats Codex at multitasking — though Codex still wins on computer use.
More from coding & agent
- Dev burns 90k credits building ambitious game with Spawn AI agent — AIandDesign · 2026-10-11
- Opus 5.5 doesn't reset prompt cache when changing reasoning effort, notes daniel_mac8 — daniel_mac8 · 2026-10-11
- daniel_mac8's blog expands on the four habits behind Asana's 89% prompt cache hit rate — daniel_mac8 · 2026-10-11
- Asana cut a Codex agent's cost 4x to $0.47 per run with an 89% prompt cache hit rate — daniel_mac8 · 2026-10-11
- Skip the orchestration layer: let your AI agents email each other — NickPassig · 2026-10-11
- Husband uses Claude Code to write a CLAUDE.md guardrail file for his wife's projects — ColleenMBrady · 2026-10-11