Copasetic Flow

Posts

Can Agents Think Outside the Box?

With all the work that's been put into making agents "correct" by construction, I gotta say, sometimes I need an LLM agent to take a chance at just being wrong. I'm working on a book project called The Gladych Files . While the book is narrative nonfiction about the history of general relativity research, it explores the liminal space inhabited by very rich fringe scientist speculators of the 1950s who funded mainstream general relativity advances, (more or less on accident.) In those spaces, you'll find Tesla, the architect of the FBI building, Timothy Leary's LSD explorations and many, many other things, institutions, and people. I've accumulated hundreds of pages of historical documents from various archives, and I'm using orchestrated agentic AI, (in the form of Gastown), to review those documents. So far, the analysis has gone well, but last week I saw something that made me look up. I'd accidentally input the same archive page twice, so i...

Gladych Files Lab Book: Document OCR vs LLM Model vs Cost or Opus is Cheaper than Sonnet for OCR!

I started my lab book entries when I was a physics graduate student. It's kind of amusing and kind of cool how far I've come. I have the equivalent of a grad student, (aka Claude Opus 4.7), working for me now. I spent some time over the weekend setting up an OCR framework for a book research project of mine. I've been coming up to speed on evals, so I decided to run one to determine which model was the most accurate and cost effective for doing OCR on travel manifest pages. I stepped the eval along rather than automating it and talked the results through with Opus as I went. First, it turns out that Opus at low effort is the most accurate and the most cost effective choice! That was a surprise. The result has to do with Opus' ability to look at higher res images which means it needs to think less for OCR vs. Sonnet. Second, at the end of the eval, as I was preparing to write up my results it occurred to me that I could ask my grad student to do it instead. Here's...

Working with Process Revision Control

I took time to play with a new Dolt enabled app example called Quorum last night. Quorum sets 13 LLM agents with different defined personas loose on a users question. The agents come up with solutions to the question and then discuss their individual solutions with each other to arrive at a consensus. There's much more detail in this blog post that accompanies the app. Quorum is cool. It is not, however, what I wanted to talk aobut here. Instead, I'm going to focus on the blog post for the app. In short, I'm very excited to see ideas that I've used to manage verification processes for years get codified into tools for LLM agents. Here's one of the important parts " I can shut down the app, lose the server, or disappear entirely — and the deliberation history remains, publicly accessible and cryptographically verified. " Imagine what an engineer can do to work back through their debug hypothesis tree with that sort of infrastructure! As the article'...

Working Through McConnell's Tensor Book

I'm working through McConnell's tensor textbook, the one I used in general relativity class at New Mexico State. I'm probably not going to be looking at spacetime a whole lot this time as I'm more interested in machine learning model implementations this time. The first video in the series is shown below. Each video has a link to the next part in the series. You can find the textbook on the Internet Archives.

Iambic Paddles capable CW Simon Via Halikey Serial

You can now play the CW Simon game on your desktop with your own iambic paddles and a Halikey Serial . (We hope to have support for Halikey Midi soon.) I was a little worried that the serial version of the Halikey was going away , but I'm super-psyched to see that it's back! I'm still pretty psyched about phone keying, so expect to see more features there, and maybe some hardware as we move forward, but there are a lot of hams with iambic keys, and so, voila! The video walks through the basic process of using a Halikey with the game. If there are any questions, suggestions, or issues, please let us know in the comments below. References More on the CW Simon game

Using ChatGPT to Tanscribe Homework

This... is interesting. The oldest kid here is taking Calc III, The Calculus of Multiple Variables. There's a whole thing every week where she winds up spending a lot of time transcribing her homework into a more readable form so that she can turn in her work on the assignment's due date. I wondered if an LLM could help with that. I tried out ChatGPT 5.3 with this picture to start. Believe it or not, this contians two homework problem, (and, accidentally, my foot.) I asked "Please write these two math solutions as laTeX and write them to a pdf file." I wasn't too surprised when it replied with --------- "I only have one photo so far, and parts of it are too hard to read reliably to typeset without guessing. From the visible top portion, I can make out a problem that looks like T ( x , y ) = 60 1 + x 2 + y 2 T(x,y)=\frac{60}{1+x^2+y^2} T ( x , y ) = 1 + x 2 + y 2 60 with partial derivatives like ∂ T ∂ x = − 120 x ( 1 + x 2 + y 2 ) 2 , ∂ T ∂ y = − 120...