Internal R&D lab · 2026

AI build orchestrator

An idea-to-deployment pipeline that researches with one model and builds with another — triggered from chat.

  • Code & DevOps
from chat message to running prototype
<1 hr
Architecture

How it fits together

  1. 01

    Idea

    • Discord
  2. 02

    Research

    • Ollama
    • Gemini
  3. 03

    Build

    • Claude
    • Codex
  4. 04

    Run

    • Node.js
    • PM2
Outcome

Ideas go from a chat message to a running prototype in under an hour.

The problem

Good ideas died in notes apps. Turning a one-line idea into something running took an evening of boilerplate, research and setup.

What I built

  • A chat bot that captures an idea and hands it to a research model for a short brief.
  • A build stage that passes the brief to a coding agent, runs it in an isolated folder and reports progress back to chat.
  • A provider ladder: local models first (free), then cloud models only when needed.

Result

Prototypes now start from a single message, and the cost of a typical run is a few cents because most of the work happens on local models.

Placeholder case study — replace with your own write-up.