@Mr_Buzzoni packs a Jev plus Opus 5.5 company brain into a 14-page PDF taken from the TypeSafe and Anthropic docs, and lists ten steps from a typed Jev answer through to judging every run.
- 6.3k views
- 134 likes
- 131 bookmarks
- 14 reposts
About this case
Guide. Mr. Buzzoni (@Mr_Buzzoni) says Opus 5.5 thinks, Jev decides, and code holds the branch. Jev returns a typed answer with a calibrated probability in 0.44s for $0.00035. Choice, Score and Noul run in parallel, and he writes that about 99% of turns end when 0.999 goes straight into an if. Routing Opus 5.5 to Sonnet and back costs 5.84 against 3.32 for staying on 5.5. Cache reads at $0.20 per million tokens are 20 times cheaper than a fresh load. The toughest 1% goes to Opus 5.5, which he credits with 1M context and 66.4% on Terminal-Bench 4.0. Judging 10,000 traces is $3.50 a day and, he says, matched the human label on all 500 decisions. He quotes the result as a loop that spends a fraction of a cent to notice and pays a frontier model only when it has to think. The personal version he names is Karpathy's second brain, with Claude keeping a wiki in Obsidian. Video attached.
Original posthttps://x.com/Mr_Buzzoni/status/2119675855651692886