i'm in SF for yc startup school! if you're an engineer, builder, researcher - hit me up, would love to chat about what you're working on

sama alpha leaks on AI-human interfaces: - end goal is ‘computers will melt away’, all interactions will be through the agent - realtime video is a singularity for UX - agents will run in the background; human intervention only when the agent asks for something

we use LLMs similarly to mainframes in the ‘70s - compute is timeshared by having a slice in the batch dimension models will compress over time, and with this we’ll be able to run more on-device thanks Andrej Karpathy for the shout out 🔥

Cost to run DeepSeek R1 (fp8) on Apple Silicon: $20,000 Cost to run DeepSeek R1 (fp8) on H100s: $300,000

i never understood 'sf is real-life twitter' until i was there. that place is networking 24/7; everyone and everything important is there. successful british founders told me, 'staying in the uk is a losing battle. you need to give in and move to sf'

clearing out llm-generated slop is the textual equivalent of tidying up the morning after

Why no popular prompt library site yet? 🤔 Prompts take time to write, but are often reusable. This can be commoditized! People often lack ideas for prompts too! A free prompt database would solve this.

Kimi K2 was released 15hrs ago and is by far the best open source model, beating DeepSeek R1 in most benchmarks 10 hours later, Sam Altman delays the release of OpenAI’s new open weight model I don’t think this is a coincidence; maybe OpenAI are back to the drawing board to make

‘The number of Macs that can train together coherently doubles every 2 months’; I'll call this 'Cheema's Law'. And it might sound a joke, but it has been remarkable how much progress we've made on this problem in such a short time. When you're working in a space that is mostly

SPARTA

‘Temperature’ is a way to control model creativity. I want a ‘context fidelity’ dial: how much a model trusts what it’s being told versus its own knowledge. When writing code, I want to specify how much the model should trust my codebase vs. how much it should believe its own

Excited to speak at Cohere Labs’ ML Efficiency meet next week. We’ll be talking about some of our recently published results about large-scale training on Apple Silicon, and how we designed an optimizer specifically with Mac training in mind. Additionally, I’ll give a teaser