Tommkruix123
Mining Architectural Quality Under Agentic AI Adoption
Building a Dependency Graph of Our Python Codebase
Show HN: AgentRuleBench, does AI agents violate inferred architecture rules?
Re:Show HN: Archprint, infer architecture lint rules from your repo's ...
on Tuesday September 08, 2026 @12:48AM (#49605694)
Yes I agree, been able to bootstrap the rules instead of writing them by hand and porting them from one repo to another. And honestly that is the limitation right now but tools like Tach does similar work for python but just does not gates each rule on statistical evidence before suggesting and also act as its own linter.
But what language do you think will worth the next real work to support?
Show HN: Archprint, infer architecture lint rules from your repo's import graph
https://github.com/Tommkruix/archprint
This tool started as a research for me trying to answer the question "Does AI agents drift from your architecture conventions or if you need enforcement instead?". So I started out building a benchmark to test that premise and on the scope I tested, the agents did not drift and I had to stop that research line. So to reuse what I have learnt from that research, I started out trying to catch...
Show HN: Archprint, infer architecture lint rules from your repo's import graph
https://github.com/Tommkruix/archprint
Every TypeScript architecture linter makes you write the rules by hand (dependency-cruiser, eslint-plugin-boundaries, Sheriff, ts-arch). Archprint does the opposite: it reads your repo's real import graph, finds the boundaries your code already respects, and turns the ones that pass a statistical confidence gate into ready-to-install lint rules, with the evidence attached (how many files...
Re:I Don't Want AI to Be a Senior Engineer
on Thursday January 22, 2026 @03:35PM (#46724799)
I built an AI agent that behaves like a junior engineer instead of a senior one.
It executes workflow deterministically: picks exactly one eligible ticket, enforces rules, routes work across repos, drafts a small PR, then hands review and judgment back to humans.
The whole system runs on GitHub Actions + APIs (no workflow SaaS). It often produces partial implementations or TODOs rather than “solutions,” by design.
This is not about replacing engineers — it’s about removing workflow friction and cognitive tax. Happy to answer questions or hear where people think this would break down.