An informal collection of promising, potentially useful agent skills, tools and information resources as I re-orient my software engineering process in the age of AI agents.
The motivation is to extend my (and my agents') awareness of what people are experimenting with and building for this new era, while keeping a prinicpled commitment to hard-won lessons from the engineers who built the world we now inhabit, many of the lessons, best practices, perspectives and strategies still hold important value, especially when confronting the problem of how to build quality, maintainable, evolvable software. I am especially attracted to experiments, projects, writings and videos that in one way or another align with this perspective.
This repo references several submodules that serve as read-only research material and AI context. No development or contribution to the submodules is expected.
Cloning (first time):
git clone --recurse-submodules git@github.com:stew/ai-swe.gitIf you already cloned without --recurse-submodules, run:
git submodule update --init --recursiveUpdating all submodules to latest:
# Fetch + checkout the latest commit on each submodule's tracked branch
git submodule update --remote --recursive
# Stage the new SHAs and commit them in this repo
git add .
git commit -m "submodules: bump to latest"
git pushEach submodule tracks its default branch (main for most, master for trafilatura),
configured in .gitmodules. The commands above fast-forward every submodule to
the tip of that branch, then record the new pins so other clones get the same
snapshot.
If you ever need to revert submodules to the pinned commits recorded in this
repo (without pulling new changes), run git submodule update --recursive
(omit --remote).
Why submodules? These external repos evolve independently. Submodules pin them at known-good revisions so that AI tools and readers get a consistent snapshot. When there's new material worth surfacing, bump the pins deliberately.