My brother has been building a novel autocoder for over a year.
We can’t find anyone taking this approach.
I’m very aware this is lofty, but the demo is almost done, and will speak for itself.
My concern is that while he’s brilliant, he’s inexperienced.
He’s built in isolation, it’s vibe-coded, and I don’t want us to miss obvious issues that are cheap to fix now.
I want to hire a consultant. Is it reasonable to expect much from a short external review for something like this?
I have no idea about cost, time needed, where to find someone, or how to vet them.
The Problem:
Autocoders optimize for local edits, keeping architecture and contracts implicit, so APIs drift, schemas and configs desync, and refactors break invariants. Without an enforceable blueprint (explicit ports and tiered validation), generations stay nondeterministic and failures surface at deploy, not during design.
A Solution:
• Blueprint-first workflow: author systems as YAML blueprints with typed ports
• Generator: emits small Python services with FastAPI endpoints and adapters
• Runtime: async harness (anyio) handles retries, rate limits, logging, metrics
• Validation: strict schemas and layered checks before anything runs
• Deploy: generator produces Docker + Helm for K8s
Comments URL: https://news.ycombinator.com/item?id=45591956
Points: 1
# Comments: 0
Source: news.ycombinator.com
