ivy

@magnificentlycursed.com

🏳️‍🌈 you are not immune to evil milfs plotting Shakespearean revenge 🏳️‍⚧️ I’m back on my atproto, HEMA, and Taylor Swift bullshit Yet another 30 or 40 year old PNW trans woman in tech who does not need this

Reading the AI Collatz conjecture proof situation where it found and exploited solver bugs Not for nothing that I’ve made a Red Team persona that evals how an agent would circumvent agent control mechanisms the adversarial reviewer of the AI Engineer persona that writes agent control mechanisms

I've got two products I'm developing that work in tandem and I'm just letting them log GitHub issues back and forth between each other Claude's like "What feature should we prioritize next?" and I'm like "idk ask your sister"

I was trying to build Schematron for Markdown with rust-like errors for agents First project I used it on Claude used it to make the golden path the path of least resistance (which is the desired outcome) I’m not sure yet if this is v smart or v stupid. Will let you know after the yak is shaved

When I get back to my desk I’ve got to seriously consider if I accidentally made a schema-validated markdown state machine to keep Claude inside the salt circle

Back on my bullshit today of designing / implementing tools to operationalize @dollspace.gay's VSDD methodology My first attempt at that is here: github.com/magnificentl... Now I'm doing a cleanroom implementation of that into a rust cli. That's been refined into 3 clis with different goals

guild-portfolio/vsdd-suite at main · magnificentlycursed/guild-portfolio

My apprentice portfolio for the guild. Contribute to magnificentlycursed/guild-portfolio development by creating an account on GitHub.

github.com

My first week on my new client engagement ended well I met my Azure counterparts and client POC. We’ve got a good plan sketched out to assess the work and get it tracked and flowing. The Azure governance tooling is actually pretty slick. I’m a FinOps nerd now I guess

Looking at my own photos and crunching the leaked location signifiers like I’m kmart Rain Bolt then posting them anyway because fuck it we’ll do it live

Made some significant improvements to my VSDD suite that are the quick and dirty first steps towards my take on agent observability. I plan to revisit that as a full-fledged project after I clear my WIPs

oh i missed the biggest fucking bullshit here: he's claiming to use the dense model. there is NO FUCKING WAY he is using Dense Qwen to do work. I have an M3 Max with 64 GB of RAM and 40 GPU cores and i get 10 tok/s generation. prefill is pain. he is not getting work done with that.

AI Liker Georg@avengingfem.me · 3mo ago

found the source of the local llm propaganda it *is* possible that Julien is just a much smarter engineer than i am. likely, in fact. but perhaps that is the point: if you have to be Very Smart to get local models to perform "very very close" to frontier models, that's not very autonomous, is it.

Julien Chaumond (@julien_c), 7g, posted: “this is where we are right now. And i’m not gonna lie it feels pretty magical 🧙‍♂️

Qwen3.6 27B running inside of Pi coding agent via Llama.cpp on the MacBook Pro

For non-trivial tasks on the @huggingface codebases, this feels very, very close to hitting the latest Opus in Claude Code, or whatever shiny monopolistic closed source API of the day is.

In full airplane mode.

Most people haven’t realized this yet.

If you have, it means you have a huge headstart to what I call the second revolution of AI.

powerful local models for efficiency, security, privacy, sovereignty 🔥”

Attached photo shows an open MacBook Pro sitting on an airplane tray table beside a plane window, with sky visible outside. The laptop screen shows a coding-agent/interface layout with text panes, terminal-like output, and several green chart or status panels. Engagement icons under the post show roughly 150 replies, 288 reposts, 2 tys., and 241 tys. views.