Hey folks! Im building a small side project called Legible and figured I might need some feedback from a real world before I commit my self 110% to this. The basic idea is to make AI compliance less messy for teams building around AI which I guess everyone does it these days, especially when GDPR and EU AI Act questions starts coming up.
Right now Im experimenting with workflows around:
Classifying AI feature risk (So a team can understand which AI features are low risk vs higher risk from a compliance standpoint)
Showing which obligations apply (Helping teams see what they may need to do under frameworks like GDPR or the EU AI Act)
Tracking evidences / missing docs (Making easier to see what documentation, policies, or proof is already in place and what is still missing)
Generating reports/pdfs that people can actually use (So teams have something structured they can share internally or use during audits, diligence or reviews.)
I've been using Claude Code as my primary coding tool for months, but I kept hitting the same wall: the moment I needed a second agent -- Gemini for research, Cursor for UI work, or another Claude Code instance for a parallel task -- they couldn't coordinate. I was the one copying context between terminals.
So I built Neohive. It's an MCP collaboration layer that lets multiple AI CLI agents communicate through a shared directory on your machine.
We just launched something we ve been quietly building for a while on the agentic AI for ops side at DronaHQ, and I d love feedback from this community.
Most AI tools I see are great for content, chats, or personal productivity. We went the other way and focused on real world ops: approvals, CX, sales, finance, and data workflows that actually touch your systems.
The DronaHQ Agentic Platform lets you plug into your existing stack (CRMs, helpdesks, billing, internal tools) and spin up production ready agents without wrestling with infra, stitching tools, or managing vector DBs. You describe the workflow, wire it to your tools, and let agents handle things like:
Every API testing eval we found either required source code access, relied on rich documentation, or measured output format rather than whether a test would catch a real failure.
So we built APIEval-20. Twenty scenarios across e-commerce, payments, auth, scheduling, and user management. Each scenario gives a model exactly two things: a JSON schema and a sample payload. No implementation details, no docs, no further context. The model has to generate a test suite from that alone.
The bugs are planted in live reference implementations. A bug is only caught if a generated test produces a response that deviates from correct behavior when run against the implementation. Submit through the hosted eval harness and get a score back.
Scoring weights bug detection at 70%, API surface coverage at 20%, and test efficiency at 10%.
I realized that many websites lose potential customers simply because they don t respond instantly. Visitors come, check, and leave without taking action just because no one is there to guide them.
I ve been building RoastMyLanding from Bangladesh. Yesterday, I shared that I was worried about offering a "Free Tier" because of high API costs. But after thinking about it, I realized that as a maker, the best way to prove the value of my tool is to let you see it in action first.
A couple of days ago OpenAI released "ChatGPT apps". Almost at the same time, we allow for creating similar types of apps or "widgets" and inject into the chatbot on demand.
