ethan1998··on Show HN: I made an agent can self-improving while workingI tried to implement RSI in my own business, but it is hard to decide when to update harness and when to update model weights.
ethan1998··on Show HN: AURA – Open-source behavioral threat detection for LLMsI need to know whether a link in a response is truly effective or if it is an ad.
ethan1998··on Show HN: Livenerf – a benchmark for whether Opus 5.5 gets nerfedI have no idea how to do it. Keep rerunning benchmarking test can be costly.
ethan1998··on Show HN: AgentRun: DSL to turn agents into workflowsInteresting, I want to know more about the comparison test with running codes directly.
ethan1998··on Show HN: Does per-step reasoning effort save money in Claude Code?If you change reasoning during a session, LLM needs to recompute KV cache. Maybe I get it wrong, but I don't think it can save money.