andrewmonostate

4 ms·

1. 1.▲
   by andrewmonostate 8mo ago

   Treni is a runtime project, not an agent framework. The problem we’re targeting is that most agents today operate on serialized tool outputs (HTTP -> JSON -> string parsing). That means the policy often cannot observe execution qual
2. 2.▲
   Show HN: Treni – single-binary GPU runtime for uncertainty-aware agents 5ms TTFT(treni-docs.pages.dev)

   2 points by andrewmonostate 8mo ago|1 comments
3. 3.▲
   by andrewmonostate 1y ago

   Thanks, please do try when you got some time! https://github.com/monostate/weave-logprobs-reasoning-loop or https://colab.research.google.com/github/monostate/weave-log...
4. 4.▲
   by andrewmonostate 1y ago

   Thanks for bringing this up! Good catch on the similarities! Yes, both use entropy/uncertainty to allocate compute intelligently. From what I understand, Entropix is an entropy-aware decoder - it monitors token entropy during generatio
5. 5.▲
   Show HN: Entropy-Guided Loop – How to make small models reason(github.com)

   33 points by andrewmonostate 1y ago|4 comments
6. 6.▲
   by andrewmonostate 1y ago

   OP here. This is a project I built during the YC hackathon. The idea was to make fine-tuning as simple as describing vibe coding, but with proper rails - no notebooks, no GPU setup, pytorch problems, venv chaos, etc, no ML background requir
7. 7.▲
   We Vibe Trained an AI(monostate.ai)

   2 points by andrewmonostate 1y ago|1 comments
8. 8.▲
   Vibe Datasetting(datasetdirector.com)

   1 points by andrewmonostate 1y ago|0 comments
9. 9.▲
   by andrewmonostate 1y ago

   My article goes on how predictive entropy, measuring uncertainty in transformer models’ token predictions—drives hallucinations in LLMs, leading to outputs like fabricated facts or events. I’m curious if broader uncertainty (e.g., epistemic