Find someone who loves you the way o3 loves putting information in little tables.

I’m going to hack on agent side projects exclusively. I need to narrow my focus to keep up.

Crazy that this is going to know you better than you know yourself

One area I’m especially interested to see for GPT‑5 is OpenAI’s pull‑request (PR) evaluation. OpenAI uses this benchmark to assess how close its models are to automating the job of an OpenAI research engineer. From the o3 and o4‑mini System Card: "Measuring if and when models

Almost no one was thinking about “agentic” AI a couple years ago.

Crazy that this tells me how to cook my food

GPT-4.5 is internally called Orion and was supposed to be named GPT-5, but unfortunately, it was not useful at all besides making greentext memes

Interesting point. GPUs per year remain the bottleneck.

imagine if gas stations didn't tell you how many gallons you were getting because car mileage was a trade secret and the gas station owned the car companies and you could either buy way overpriced gas per-mile or a monthly "max gas subscription" that turns off randomly sometimes

AMD MI300X tokens/sec beats H100

For folks training/fine-tuning small-ish models, what’s your environment of choice? Do you use notebooks or scripts?

If you train/fine-tune models, I'd love to chat with you. Please comment or DM and I'll reach out!