jbellis
Joined 19 years ago
Joined 19 years ago
Here's mine for coding in Rust, and I definitely see the same pattern of "most chinese models do much worse than on public evals that have been around long enough to train on" that AA v5 surfaces: https://slopcop.com/power-ranking
(And yes, I am using the word "ban" as shorthand for "extremely tight/onerous regulations in red and blue states, respectively".)
Reasonable overview here and it's only gotten worse since 2024. https://thezvi.substack.com/p/the-online-sports-gambling-exp...
two things help:
1. fs-level compression, I use ZFS bc Btrfs has massive contention problems with concurrent Rust builds
2. Archive your old sessions using something like sessionwiki (https://github.com/youdie006/sessionwiki), I've integrated this into my control plane (https://github.com/BrokkAi/mjolnir/)
And yes, Opus is enough smarter than DSF that it's worth the extra steps. This ranking is from live tickets, no contamination: https://slopcop.com/power-ranking
My harness control plane will autoconfigure mbx across all your machines, if you like. https://github.com/BrokkAi/mjolnir/
Brokk keeps LLMs on-task in million-line codebases by adding compiler-grade understanding of your code's structure and semantics.
Previously: author of JVector, co-founder of DataStax, founding project chair of Apache Cassandra.
Twitter: http://twitter.com/spyced