How to use Codex to prototype a retrofit lab valve
Codex helped a petroleum engineer prototype a retrofit ESP32 valve actuator, with BLE control, stop-on-release behavior, physical tests, and manual backup.
devpost.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Help prototype a retrofit actuator for this existing lab valve without removing manual access. Review servo torque and motion limits, ESP32 wiring, BLE control, and Arduino firmware. Implement hold-to-run, stop-on-release, and reviewed presets. Produce a test plan and wait for my physical validation; do not claim autonomous high-pressure operation.
Reach people exploring AI agents
See placements →What happened
A 3D-printed retrofit valve-control prototype was built and tested with ESP32/BLE servo firmware; broader sensing and autonomous experiment planning remained future work.
Reported by James Wang — DevpostHow the agent did it
- James specified a retrofit requirement that preserved existing pressure-rated valves and backup access.
- Codex helped design ESP32, BLE, and servo control with hold-to-run and stop-on-release behavior.
- James printed the mechanism, reviewed implementation outputs, and tested physical valve-control behavior.
Where a person stepped in
James supplied petroleum-lab requirements, selected the safety boundary, built and tested the physical prototype, and retained manual or robotic takeover.
Petroleum engineer James Wang wanted to automate valve handling in enhanced-oil-recovery laboratory experiments. Existing pressure-rated rigs were too valuable to replace wholesale, so he used Codex to work toward a retrofit actuator rather than an entirely new laboratory system.
Codex helped reason through the mechanism, servo behavior, controller choices, wiring, Bluetooth control, and Arduino-ready firmware. The prototype used an ESP32 and a FashionStar serial-bus servo. Its control model included rotation while a button was held, stopping on release, and defined angular presets.
Wang built and tested a printed valve-control mechanism. The design preserved the existing valve stem for manual or robotic backup, and he reviewed outputs from parallel Codex work before trying them on the physical prototype. The account presents that functioning control layer as the accomplishment.
The larger autonomous lab remained future work. Pressure, temperature, and flow sensing, gauge interpretation, and higher-level experiment planning were still planned integrations. Those distinctions matter in a hazardous application: a demonstrated actuator and firmware path do not establish a certified high-pressure experiment controller. The useful workflow was translating domain constraints into a testable physical control model while retaining human authorization and takeover.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Devpost · James Wang
Frequently asked questions
Did the project replace the laboratory valves?
No. Its actuator was designed to retrofit existing manual valves.
What happens when the control is released?
The reported control model stops rotation and holds position.
Was full autonomous experimentation demonstrated?
No. Sensing and higher-level planning were listed as next steps.
More details
About this account
James specified a retrofit requirement that preserved existing pressure-rated valves and backup access. Codex helped design ESP32, BLE, and servo control with hold-to-run and stop-on-release behavior.
Firsthand petroleum engineer’s prototype account. Physical actuator testing is reported; future sensing and AI-planned experiments must not be presented as completed. No certification or unattended hazardous operation is evidenced. Suggested prompt written by Agents In Use; no original task prompt was published in the verified account.
Credit: James Wang — Devpost
Source published 20 Jul 2026
Source checked 14 Oct 2026
Tools and topics
Codex · ESP32 · BLE · FashionStar servo · Arduino IDE
valve control · retrofit hardware · manual takeover
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.