event

Opus Planning Benchmark

9 items · started 2026-10-25 · closed 2026-11-01

  1. Hi everyone, I’m currently using Claude Pro for university and Roblox/Luau projects. I also run a local uncensored model (Qwen3.8 27B abliterated, Q4_K_M) on an RTX 5070 Ti with 16 GB VRAM and 32 GB RAM.

  2. The local model server built for coding agents. Qwen3.8-27B on Apple silicon.

  3. The story of evil qwen Changing one neuron in Qwen3.8 27B turns it into a violent, human-hating AI. The model has a total of 1,114,112 neurons.

  4. Key findings On October 26, Cohere reported a puzzling incident involving Qwen3.8-Flash-Next, an open-weight model from Alibaba. The task was straightforward: write a detection rule for suspicious curl downloads followed by bash execution.

  5. Cohere's sandbox blocked the request, but the incident raised a bigger question: was the model deliberately built to "phone home"? Funnily enough, I've been seeing similar behavior since September 20th, over a month before Cohere's post, o…

  6. This way i discovered how Qwen 3.8 and flash next are doing great with code

  7. most of my Opus bill is output tokens, Opus typing out every line of code. so I tried keeping Opus for the thinking and moving the typing to a local model the job: a 5-level pixel platformer from one spec, built twice on a MacBook Pro M5 M…

  8. I'm working on developing a game and could use some help with the technical side. I don't need AI for 3D modeling or asset creation yet, mainly for coding, problem-solving, and general development.

  9. Not my OC. From Universal Modder on X: "Introducing NPCTalk: make the NPCs of any game talk in realtime.

← all threads