How are you checking whether your agents work? (not just one test based on vibes)
So I think I'm turning crazy. Everyone has a crazy agent setup on twitter. But I personally keep running into the same wall building agents. I…
Latest
CTX: How far can you reasonably go with Qwen 3.6 27B?
How far can i stretch the context window with Qwen 3.6 27B (using Q8_0) before it gets too unreliable? I am at 100k right now and…
Dual 3060 MoE loading issue.
I had one RTX3060 and was able to use all models like Qwen3.6 27B, Qwen3.6 35B and Qwen3.5 122B with CPU offload. I've added second 3060…
I created a super harmful model ! :D (by tweaking it’s J-Space!!!)
Soooo! Since Anthropic share their Jacobian-Lens a few days ago I went on and made a tool based on it which adds the possibilité to export…
I feel like I’m not using my hardware efficiently
Hi there, got a 7950x,128GB DDR5, RTX 4090 and RTX 3090TI. I'm currently running Qwen3.6 27B Q8 with 262k Context at Q8 with llama.cpp. It's not…
6x MI50’s on PCIE vs 4x MI50’s on PEX8749 and 2x on PCIE
I am really excited to share this one. On the X99-E-WS motherboard.. while old and PCIE 3.0 - I think it's still pretty capable for what…
Follow-up: what I learned scaling a SQLite/FTS5 patent database from 3.5M to 5.36M records
A few months ago I posted about classifying 3.5M US patents with Nemotron 9B on a single RTX 5090. This is a follow-up on the data-engineering…
Why are MoE models so belittled?
E.g "Qwen 3.5 122B is just 10B active, so it's no where close to the dense 27B model" That is the main sentiment around here and…
Are EPYC CCDs all you need + benchmarks
Recently I've found a deal to buy 9374f for cheap to replace my bottle-necked 9135. 8 CCD looked delicious. But first benchmarks showed me no decoding…
Built EverFern, an alternative to cowork open-source(LangGraph + Electron). Tested it against Qwen3-8B, looking for feedback
I built EverFern, an open-source (MIT) desktop agent that does computer use, browser control, and file/code tasks, aimed at people who don't want to pay $20–200/mo…