Building specialist AI datasets • Fine-tuning & evaluating LLMs • ML student • Glassblower by trade, Sharing my workflows, also Hardware projects
Norway
Joined September 2026
- Tweets3.4K
- Following250
- Followers282
- Likes4.5K
I will, I don't benchmark with standard benchmarks, I hook them to my PI agent with subagents built just like my subagents in Claude. They shadow frontier in my real work flow and are graded pass or fail. this model will be tested in the coder position, If he passes, That's my largest daily expense right now. I will give you the feed back, I can save all the tool call traces if you like.
This is the feedback so far. Update on converting my wife to Omarchy, since dhh asked what tripped her up:
She wanted it to feel like Windows, so I added floating windows and a taskbar.
She didn't want a frontier model, so she got Hermes running Strata Qwen Flash Next locally.
She danced in her chair.
I told you bio tech is where these companies will hit the big money.
GPT-Rosalind surfaced a new lead for drug discovery: could restoring cholesterol transport protect neurons from degeneration?
DCTN1 is a gene linked to Perry syndrome, a rare neurodegenerative disorder. In mouse-brain data, disrupting it increased expression of genes involved in making cholesterol. GPT-Rosalind suggested a possible explanation: cholesterol may not be reaching where neurons need it.
That gives scientists a specific mechanism to test and, if confirmed, a direction for developing new treatments for neurodegenerative disorders.
With @perturbai_tx and @scrippsresearch, a team of thousands of GPT-Rosalind agents worked in parallel, connecting gene-expression results from millions of mouse brain cells with scientific literature in just 32 minutes. We’re now expanding this collaboration to identify new drug targets.
openai.com/index/scaling-ai-…
Local is getting very interesting these days, I need to get this one downloaded.
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Now, I hooked her up on Hermes running local on Qwen Flash Next on the Strata engine, the largest one. I had to windows it up a bit for her, she really needed to get things done and got a little frustrated, last I looked she was dancing a bit in her chair.
I would run them in a real work flow shadowing frontier models then grade them pass or fail after review.