Infer an undocumented device protocol from captured traffic and turn the confirmed message structure into a small control tool.
Source episode · 16:08
How I AI: My GPT-5.5 Review—A 6-Hour Autonomous Task and the Bluetooth Hack No Other Model Could Solvewith Claire Vo
16:08 to 22:30: Claire combines packet captures, protocol notes, and failed attempts, then asks Codex to infer a proprietary Bluetooth transport and build a working controller.
A documented protocol map and a test tool that reproduces selected device actions safely.
OpenAI's cloud-based AI software engineering agent that can execute code, run tests, and handle complex multi-file tasks autonomously.
Step by step
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
5 steps
Use a packet sniffer and a developer profile on a mobile device to intercept and log the raw data being communicated between an app and the target hardware (e.g., a Bluetooth speaker).
Capture repeated packet traces for one known action at a time and label the exact user action and time.
Gather all available information into one place: the captured packet logs, your personal notes, and any obscure documentation you can find, regardless of language or clarity.
Combine captures, device details, public documents, and failed hypotheses into one indexed research bundle.
In an advanced coding environment like Codex, paste the entire collection of your research into the context window. Include the raw packet logs, your personal notes, and any documentation you gathered, no matter how fragmented or obscure.
Map repeated bytes, variable fields, checksums, message direction, and timing across the labeled captures.
Prompt the model with your high-level goal. Explain what you want to achieve with the hardware and instruct the AI to use all the provided information to figure out how. For complex, undefined problems, it can be effective to grant the model broad latitude to find a solution.
Infer the smallest protocol needed for [target actions], show confidence and alternatives, and propose a safe distinguishing test.
The model should return an analysis of the data that deduces the proprietary protocol. Review its findings, then work with it iteratively to translate this understanding into a functional script or command-line tool that can programmatically control the hardware.
Build a local tool that sends only the confirmed commands, logs raw responses, and refuses unknown or unsafe payloads.
Turn an idea into a PRD, user stories, and a plan.
After the steps
How to recover when the loop fails and where human judgment helps.
Recover
Keep building
Use a hybrid AI workflow to bridge design and development. Generate high-level design concepts and tokens with Claude, then use Codex for robust implementation of the technical components in your codebase.
Implement a QA loop for AI-generated code using Devin. This playbook combines automated code reviews with an agent-recorded video walkthrough for visual verification, ensuring high-quality merges at scale.
Build a reusable 'Watchdog' playbook in Devin to automatically monitor customer accounts. The agent proactively checks for errors, synthesizes a prioritized list of critical issues, and saves you time.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.