live
- your daily tech & design digest, one email a day -
PyTorch's ExecuTorch brings GPU inference to the browser via WebGPU [Xtra!]On-device/in-browser ML inference without a server round-trip gets meaningfully faster.
Goodfire maps neurons as an interpretable basis for model internals [Xtra!]Interpretability work like this is foundational for trusting and debugging frontier models.
AI Futures Project: intelligence explosion timeline moves to 2027 [Xtra!]A detailed forecast update that compresses expected timelines for transformative AI by years.
Google's Antigravity ships Swarm for multi-agent verification [Xtra!]Multi-agent orchestration with built-in verification is the next layer coding agents need.
Apple shows a recipe for distilling attention models into Mamba [Xtra!]Cross-architecture distillation could let teams get Mamba's efficiency without retraining from scratch.
Nvidia board reportedly moving to oust Jensen Huang [Xtra!]A potential CEO change at the company anchoring the entire AI hardware boom is a market-moving event.