anulum

4 ms·

1. 1.▲
   by anulum 7mo ago

   https://anulum.github.io/director-ai/benchmarks/
2. 2.▲
   by anulum 7mo ago

   @soletta — you're right, and I've deferred this three times now, which isn't useful.
3. 3.▲
   by anulum 7mo ago

   Hey HN — huge thanks for the thoughtful comments yesterday! I shipped *v1.2.0* overnight with everything you asked for: • Full end-to-end benchmark notebook (600+ real RAG/agent traces, HaluEval + TruthfulQA, head-to-head vs Claude sel
4. 4.▲
   by anulum 7mo ago

   @soletta Got it — thanks for the extra clarity, that’s an important distinction. You’re absolutely right: modern frontier models (Claude 3.5/Opus-class, GPT-4o, etc.) have become extremely good at maintaining internal consistency dur
5. 5.▲
   by anulum 7mo ago

   @soletta Great question — this is exactly why we built it this way. *Short answer*: frontier LLMs are excellent at static self-critique, but terrible for *real-time token-by-token streaming guardrails* because of latency, cost, and lack o
6. 6.▲
   Show HN: Director-AI – token-level NLI+RAG(github.com)

   2 points by anulum 7mo ago|7 comments
7. 7.▲
   Show HN: I just shipped the canonical neuro-symbolic control demo(github.com)

   1 points by anulum 7mo ago|0 comments
8. 8.▲
   Show HN: SC-NeuroCore – Rust neuromorphic compiler, 512× speedup(github.com)

   2 points by anulum 8mo ago|0 comments
9. 9.▲
   Show HN: SCPN Fusion Core – Tokamak plasma SIM and neuromorphic SNN control(github.com)

   2 points by anulum 8mo ago|0 comments