Hi all,

I hope everyone is well. As new AI-Agent coding tools and models are
released and improved, I've started to adopt them in my professional life
to some extent personally as well. I've tried out Claude Code and OpenAI's
GPT Codex, with a handful of models on both sides.

I can't speak to the code quality, but this is what i've done so far.

1. Tried to use Claude to optimize Text Sweeper. We iterated on many
approaches, but so far didn't find much success. It tended not to be so
good at hybrid Asm with BASIC.
2. Used claude to try and replicate my release workflow (today I was using
Virtual T to Tokenize and ROM2/Cleuseu to Pack the program down to the
smallest possible size.
3. Created a .py script to replicate the Cleuseu packing, and another .py
to do the tokenization. (Shout out to https://github.com/hackerb9/tokenize,
whose reference doc on the Tokenized file format was very useful to Claude)
NOTE: I think the tokenizer is actually still dropping comments - but my
LLM tokens ran out so I'll have him fix it later)
4. Then my tokens ran out and I switched to GPT Codex for the rest. I
basically had it create a GitHub action workflow that is triggered by
creating a release in Github. It runs the packer and tokenizer script and
then attaches the outputs to the Github release as artifacts,


TextSweeper
https://github.com/Grimakis/TextSweeper

Utils (used as a submodule in TextSweeper project for the release workflow)
https://github.com/Grimakis/model100-basic-tools-python

Please feel free to try out the Utils with your projects and let me know if
there is anything else incorrect, if you find it helpful, or otherwise.

Best,
George

Reply via email to