HN Explorer

How are you centralising AI for development at your company?

1 points by simianwords ↗ HN
I'm asking this because I'm seeing some frustrating patterns and if I may be bold, I think my answer is the correct one.

Here's one I'm seeing: its the SDLC agent pattern and I think its wrong.

Step 1: you have good AGENTS.md documentation for your code and other principles. (this is good)

Step 2: There's a somewhat strict workflow and steps enforced like planner agent, reviewer agent, critique agent and directly create a PR from a ticket. Every human intervention here is considered an anti-pattern.

I'd like it to be as simple as having documentation and giving developer full freedom with choice of harness, tools (MCP) and just make the pull request.

What's the logic behind the multi step planner, reviewer, critique agents?

4 comments

[ 1.8 ms ] story [ 23.8 ms ] thread
Xantier
I built an internal tool that does this. Having the ability to spin up sandboxes on demand to run a harness(es, depending on the task) and having a collaborative thread which multiple people can chip in with their opinions. Mostly the attempt is to make a coding harness part of the discussions that we as humans usually have during the day (on Slack, ticketing systems, GitHub etc.) and trigger the harness to do the grunt work based on those discussions. First creating a plan what is iterated on and then doing the actual coding based on that approved/finalized plan.
oscarestudillom
we are working just like you described. AGENTS.md on every project, some skills that we maintain that we use as workflow (/create-tasks /plan /implement). About the harness we all use claude code, because it's easier for billing to just have one invoice and everyone using the same. We use Revix AI as reviewer in pull requests.

The skills are purely guided questions to force the dev to answer some questions that guide the agent on how to do the things.