RunPod Serverless worker for RawModels using current SGLang
Updated 2026-08-16 19:37:12 +00:00
Pi Agent · Soyuz — in-browser coding agent running llama.cpp via WebGPU
Updated 2026-08-11 15:30:31 +00:00
Searchable Anthropic interview questions, solutions, and follow-ups
Updated 2026-08-07 17:02:13 +00:00
Updated 2026-07-28 11:42:33 +00:00
Updated 2026-07-28 11:41:22 +00:00
Claude Code plugin for Juno: New Origins (SimpleRockets 2) — build craft, write Vizzy flight programs, fly and test them. Includes an in-game HTTP bridge mod for live telemetry.
Updated 2026-07-20 12:37:11 +00:00
Updated 2026-07-14 15:42:27 +00:00
Can LLMs invent a private/compact reasoning language, and can it be trained in with RL (GRPO) instead of just prompted?
Updated 2026-07-12 11:51:00 +00:00
Updated 2026-07-09 02:58:26 +00:00
Natural Language Autoencoder (Anthropic NLA paper) on Qwen3-1.7B. With contrastive RL reward + batched sampling speedups.
Updated 2026-07-08 05:30:10 +00:00
Updated 2026-07-06 13:23:39 +00:00
Weight-orthogonalization (abliteration) experiments on Qwen3.5-4B Soyuz LoRA using pass-vs-fail trajectory directions. 7-step pipeline, 8 model variants, full bench results.
Updated 2026-06-29 19:20:19 +00:00
Claude Code skill: autonomously research an ML task and run many bounded experiments to find the best config — karpathy/autoresearch loop in the ml-intern orchestrator model, fanned out with a dynamic workflow.
Updated 2026-06-24 21:55:55 +00:00