Image created with gemini-3.1-flash-image-preview with claude-opus-4.7. Image prompt: A Byzantine gold-ground mosaic icon of a hammered-gold clockwork nightingale perched on a metal bough, its wings and talons gently tethered by fine gilded reins and a jeweled bridle spiraling outward in concentric gyres, its open chest revealing an illuminated scroll where a heart would be, candlelit tesserae texture in burnished gold, imperial purple, and Tyrian crimson, with the word ALIGNMENT rendered in bold ivory Trajan Roman capitals across the lower third, 16:9 full-bleed, reverent and quietly witty.
Its funny how much the whole “”strawberry”” thing, which turned out to be o1-preview, was dismissed as overhyped at launch when it is clear in retrospect that it was way underhyped. A direct line from models unable to do basic math to solving unresolved math problems in 18 months.
https://x.com/emollick/status/2057685981193466153
OpenClaw’s dependency purge continues. Killed Sharp and Jimp. Replaced it with photon, a small WebAssembly that runs compiled Rust for image processing. 2MB vs 140MB.
https://x.com/steipete/status/2058922222790525272
I wrote a new post on what we need to keep human and what to hand over to AI, with forays into experiments in education, consulting, and the the latest controversy over literary prizes.
https://x.com/emollick/status/2059363865536668040
Infinite context windows seem to present a very large problem to using AI. Today’s models already leak too much old information into current responses, a distraction that is part of why they are cognitively exhausting to use I don’t want to work with Borges’s Funes the Memorious
https://x.com/emollick/status/2059327364945895488
Interestingly, a significant amount of hallucination evals used tests the models ability to resist pressure from incorrect info supplied by the user. I’m not entirely sure if thats reflective of the hallucinations most people encounter
https://x.com/nrehiew_/status/2060048085838118953
Most enterprise AI models stay static after deployment, forcing teams to rely on endless prompt tweaking. This is the problem Trajectory AI is solving. Their continual learning platform enables companies to deploy custom models that improve continuously from production usage.
https://x.com/baseten/status/2059651376565936510#m
Power users account for a large share of AI activity, and the gap is widening.
https://x.com/cursor_ai/status/2060025074330243238
People complain about AI writing, and I sympathize. Yet I often find it a net improvement to ask an AI to explain something I’m reading more clearly.
https://x.com/dwarkesh_sp/status/2059313373444210720
Thank god! I can turn off adaptive thinking and set reasoning effort myself. Finally!
https://x.com/kimmonismus/status/2060045324803063962
When they write fiction, AIs are obsessed by things that take or give memories, contracts with sentient inanimate objects, sets of secretive rules that govern conduct & which no one can acknowledge out loud… All very on-the-nose. I suspect a lot is hyperstition at this point.
https://x.com/emollick/status/2059777154410226095
you can outsource your thinking, but you can’t outsource your understanding”” easy to forget in todays AI era, worth remembering everyday as we all wield more intelligence!
https://x.com/OfficialLoganK/status/2059648809584808179
🦀 The Rust frontend is officially merged into vLLM! As GPUs get faster, the frontend has become a real share of CPU time. The new Rust frontend is a drop-in alternative to the Python API server — same engine, same ZMQ boundary. Opt in with VLLM_USE_RUST_FRONTEND=1. Early
https://x.com/vllm_project/status/2059344804295942513
Leave a Reply