AI
I did not plan to write about AI, I didn’t plan to write at all, and literally removed this blog at some point, because I had nothing interesting to say.
But, “AI-first"ers got me, so I restored the blog, and oh boy, now I do have what to say.
Most of what I do with models is unglamorous: prompting them like it matters, running the small ones on hardware I own, and keeping agents on a leash - yada-yada, AI slop.
Nope. Handwritten posts from a pretty annoyed backend engineer.
And yes, I do know what your “AI girlfriend” is, because I had to spend a bit of time on learning that stuff, both from an academic standpoint (can I confess my love to Stanford University for their online courses and arXiv papers?) and from a practical side of things (when you work in billing, AI stuff like Sift is in your toolchain by default).
Posts RSS
-
'Write AGENTS.md' Skill
An agent skill to generate AGENTS.md, and why the output is, at best, a draft.
-
If It Is So Dangerous, Then Sandbox It!
Step-by-step guide to sandboxing your agent; a bunker and a local data center are optional.
-
Learn How To Prompt Already
You still have to learn how to prompt a model, otherwise who else will show it what good looks like?
-
Krea 2 Prompting Skill
An agent skill for writing Krea 2 Turbo prompts, built from hundreds of local image-generation experiments. It covers everything from photorealism to adult content and pure organic monsters.
-
Stop Enter-ing, YOLO It
A few sentences of prompting and then hours of <Enter> pressing, what a wonderful day (no)! YOLO it for the god's sake.
-
You Don't Need A %Frontier LLM%
In most of the cases you don't need (Astra|Fable|Opus|%paste-a-new-LLM-name-released-this-week%).