Loading search...
Thoughts on data engineering, GenAI, distributed systems, and software architecture. Views expressed here are my own and do not represent my employer. RSS
I run several coding agents every day. Hermes is different because its job is not to finish one coding task. Its job is to remember, schedule, and route the work that fills the rest of my day.
Developer Twitter claims you can deslop AI output by pasting ASD-STE100 or Orwell's rules into system prompts. Empirical benchmarks prove the opposite: forcing simple language at reasoning time cripples the model.
Once you run several coding agents at once, plain terminals and classic multiplexers stop being enough. Herdr fills the gap between tmux and agent apps.
I spent a weekend with Moonshot's Kimi K3 launch notes and tech report. The useful story is not leaderboard rank. It is systems: sparse MoE, long context, open weights, and what it takes to run a 3T-class agent brain.
Most coding agents treat a subagent as a nested chat that returns prose. omp gives workers typed results, direct peer messages, and safer shared-file edits.