Meta Muse hands-on: fails as tasks pile up, "scales to millions but not personally"
deliprao · x · 2026-09-19
Delip Rao ran controlled experiments on Meta's Muse agent and concluded it "scales to millions of users but does not scale personally":
- Incrementally adding complexity showed that the more cron jobs it tracks, the more slip-ups occur, until it stops following earlier important instructions;
- Beyond typical model/harness failure modes, he suspects silent failures invisible to users, like rising cron job failure rates as user count scales;
- Muse repeatedly asked for information already in its context (model issue) or in longer history (harness issue).
His take: Muse is well-designed with good UX but remains a reactive agent requiring user-initiated push, not a personalized ambient intelligence companion—perhaps not Meta's goal anyway. He draws a broader lesson about owning your own intelligence infrastructure with dedicated compute. He notes it's an n=1 observation, but argues vibe evals matter for AI products.
More from coding & agent
- Anthropic finally adds AGENTS.md support to Claude Code — iannuttall · 2026-09-19
- Stop hand-writing agent 'assembly': compilers like DSPy are the key idea — lateinteraction · 2026-09-19
- Dev builds OCR+Jev image classifier that sorted ~900 images in 40 seconds — gaganghotra_ · 2026-09-19
- Blender scenes too sparse? Iterating references with GPT Image for 14 rounds adds detail — kevinkern · 2026-09-19
- Ambient agents are finally possible, developer declares — kieranklaassen · 2026-09-19
- Trylle ships Magic Sessions: run long cloud coding tasks from issue to PR — ayushtweetshere · 2026-09-19