Recohut Field Notes
We read the research, trace it back to primary sources, and explain what engineering teams can actually apply. Every note separates published evidence from our practical interpretation.
Latest
Original Recohut analysis with clear attribution to the work that inspired it.
How agents learn from experience, with a furniture quotation walkthrough showing harness changes, response validation and technical code diffs.
Inspired by the RRSI research team for Google Research
Build an agent step by step: choose its goal, loop, model, tools, knowledge, skills, context, persistence, and verification, then inspect the code each piece adds.
Inspired by Yao et al. for ICLR 2023
A practical look at TypeSafe AI’s System One model, its typed decision contract, and how teams can evaluate bounded automation without confusing schema safety with correctness.
Inspired by Diogo Almeida for TypeSafe AI
A practical map of supervised demonstrations, preference data, reward modeling, reinforcement learning, and the evaluation loop around them.
Inspired by Sharon Zhou for O’Reilly Radar