I wish more people were paying more attention to the projects that ship and have users. With the amount of change coming to our industry, spending any time listening to exposition about some experimental thing no one uses is an inexcusable waste of time. This goes double for debating skeptics.

Ed@ed3d.net · 2d ago

I have not been "iterating" on a thing that produces no useful output for months. Instead I have been working on @polytoken.dev while also using it for other productive work (and so have other people). It's embarrassing. It's indicative of a bad state that this is considered meaningful thought.

Currently running GLM5.2 at almost 1.7 tps and don't believe I'm at the ceiling yet.

youtu.be/QyrQeeDZUG4

George Hotz AMD Advancing AI 2026

YouTube video by Dom

youtu.be

My great weakness is that I value my own time at zero and I price goods at their component cost

I wasted days fighting with LiteLLM and mlx-lm. Frontier models definitely still needed some percent of the problem space

gonna start talking like this to my boss

one shots are for two pump chumps

i can respect that some people are getting quite far in accomplishing impressive one shots, but this is starting to look like a parlor trick in comparison to building something deliberately iterative. once you open the door to orchestration, much much bigger things are possible

We don't sell vapes

I've "wasted" a week organizing ancient archives of old code and half forgotten projects, solely so i can lay my head on the pillow at night knowing i'm an organized person

Im not so much on the loop as on some loops. Some largely disposable hastily constructed loops with little reuse

I now have more free VRAM than most people have VRAM.

Canceled Cursor today. Need to figure out what to do with my Anthropic credit.

the demos are coming

In some sense, agentic programming is like employing a liar. You have to figure out how to make that liar useful even if they cannot be trusted, which is probably what it is like to design a prison or a work camp.

i've been busy and forgot to post. migrating to local LLM coding assistance has been... interesting opencode talks to litellm which talks to openrouter or llama.cpp i love GGUF but llama.cpp does so much less for you than Ollama

cross compilation is generating illegal instructions so i'm on a 2 week tangent to fix that leak

whisper.cpp rebuilt on nvidia API turned minutes of transcription into seconds

btop, but with terminal limits on the linux console no gpu on the machine with GPU