dspy-wsp follow-up: MLX/PyTorch/vLLM backends, shines on decision models like d1-2B
dbreunig · x · 2026-11-14
Author follow-up on dspy-wsp: it builds on Rikiyatakehi's weight-space programming work and Haoze He's application to decision models. It runs with MLX, PyTorch, and vLLM backends, has been tested modifying Qwen3.5-0.8B, and works best on decision models like Liquid AI's d1-2B and d1-omni.
More from coding & agent
- Perplexity Teammates screenshot shows agents on Datadog duty, reading PRs overnight — testingcatalog · 2026-11-14
- Perplexity Teammates agents roll out on Mac app for Max subscribers — testingcatalog · 2026-11-14
- Nexus launches as an agentic workforce for AI deployment teams — dr_cintas · 2026-11-14
- Nous Research's Hermes Agent Desktop gets a live aurora landscape theme — Teknium · 2026-11-14
- Users now talk to Claude mid-call and run a /post-meeting-reconciliation skill right after — Flomerboy · 2026-11-14
- LMMs Lab open-sources The Harness, an MIT-licensed autonomous agent framework — liuziwei7 · 2026-11-14