jevals
Run dozens of agent evals per trace in one sub-second request, no judge prompt
Emerged finds new products worth a look, daily — what’s new today
AI·rete·RAG
Auditable rule-engine decisions, with an LLM that only writes the explanation
For anyone automating lending, fraud or triage calls, it keeps the verdict deterministic and replayable while still giving auditors a readable reason.
A hosted decision service built on a pure-Python Rete engine: rules in YAML or a visual editor evaluate your facts and produce the verdict, with forward chaining and a full trace of which conditions fired and which did not. Retrieval over your own policy documents then feeds an LLM that explains the decision with citations but cannot change it. Policy documents can be turned into draft rules for review. There is a no-signup demo across eight domains, a free tier with paid Builder and Pro plans, and an MIT-licensed MCP client; the engine itself is closed.
More like this
CTRLRun
An open-source safety layer that checks every AI agent action before it runs
Bespoke Nimble
An open 9B model, dataset and training recipe for fast typed decisions over text
CodeEraser
A deterministic gate against duplicate code and stale docs from LLM agents
C
Concepts
Nineteen ML and search explainers you learn by dragging the inputs around
TrackLLM
A public record of when hosted language models quietly change behaviour