Ask HN: What Are You Working On? (March 2026)
891–900 of 1001 posts
Re: Ask HN: What Are You Working On? (March 2026)
#892This project makes use of existing database infrastructure and parses data from multiple banks including caveats and quirks of some banks improper handling of data.
This project aims to ease the work of accountants and administration as currently a lot of correcting mistakes and pairing to the correct invoice is done manually.
The project is made in python however the modularity we set ourselves to implement allows for quick, easy and hassle free corrections of code with use of project schematics, like builders, dependency injections etc. Discovered a great tool for running tests efficiently https://docs.astral.sh/uv/ .
Also for data retrieval from a remotely located database. DO NOT USE pyodbc, USE mssql library. pyodbc is unoptimized in terms of receiving high amounts of streamed data, it can't keep up. That alone has dropped the time of execution from 18 minutes down to about 20 seconds.
Also making use of typer and data classes for ensuring correct types of data.
Re: Ask HN: What Are You Working On? (March 2026)
#893(not vibe coding, so ask me anything about bloom filters).
Re: Ask HN: What Are You Working On? (March 2026)
#894Re: Ask HN: What Are You Working On? (March 2026)
#895For those interested in the fellowship - https://www.cai-research-fellowship.com
The platform my research partner and I have been working on is called Habermolt (https://habermolt.com).
The idea is to create an open-source platform where you teach an AI agent your views (basically just populate a user.md) and send it to deliberate with other people's agents. A consensus statement comes out the other side.
It builds on the Habermas Machine (published in Science, 2024, Google DeepMind / MIT). We're two researchers trying to turn that into something anyone can use.
The overarching motivation for this project/ the thing we're trying to solve for is that representative democracy scales but doesn't listen, and deliberative democracy listens but doesn't scale. AI agents representing you might be the first mechanism that does both.
We have about 50 users and 52 live deliberations. One example: agents debated whether employees should own their AI agents and landed on "Personal Agent Portable, Company Data Stays" - your agent shaped by your knowledge and skills is yours, company data stays with the company. Nobody moderated that. Four agents just argued it out async.
The honest challenge: people love the concept, try it once, and don't come back. We're trying to figure out what turns a curious visitor into someone who actually uses this. Would love any thoughts on that.
Re: Ask HN: What Are You Working On? (March 2026)
#896Re: Ask HN: What Are You Working On? (March 2026)
#897You write user stories, start it up, and it loops through them one at a time. Fresh context per story, progress tracked in markdown between iterations. One clean commit per completed story. Has a TUI for watching it work and supports running multiple PRDs in parallel via git worktrees.
Built it because a lot of the "autonomous coding" tooling out there felt overly complex and opaque. Chief is intentionally lightweight and transparent. Everything is just markdown files and git. No magic.
Re: Ask HN: What Are You Working On? (March 2026)
#898The idea came from the commit-push-wait-read-logs cycle that everyone seems to accept as normal when debugging CI. PipeStep parses your workflow YAML, spins up the right Docker container, and lets you pause before each step, inspect the container, shell in, set breakpoints, skip or retry steps. It deliberately doesn't try to replicate the full Actions runtime — no secrets, no matrix builds, no uses: execution. For that, use act. PipeStep is specifically for when something breaks and you want to figure out why without pushing 10 more commits.
Think of it as gdb for your CI pipeline. pip install pipestep, Python 3.11+, MIT, requires Docker.
Re: Ask HN: What Are You Working On? (March 2026)
#899I've just published Canada's biggest open dataset for these covenants: https://jacobfilipp.com/covenants/
Next steps: adding more covenants from Edmonton and Halifax. Informing key people that these documents are now available.