Key points
- The system uses multiple devices to listen and estimate locations via sound.
- A manual rough-mode adjustment improved initial positioning accuracy.
- The tool located a carbon monoxide detector within 40cm of its true position.
Simon Willison reported building a software system called Beep Locator using Claude Code to find an intermittent beep in a cluttered room.reported The tool aims to triangulate sound sources without manual searching through boxes.An engineer used an AI tool to build a system that finds hidden beeping sounds using phones and tablets.An AI agent generated a multi-device acoustic triangulation system that required manual spatial calibration to achieve sub-meter accuracy.
What was released or claimed
Willison described the system as a Python web server with a primary UI. Users connect devices via QR code or URL to act as listeners. He used two iPads, a phone, a MacBook Pro and a Surface Ultra for testing.
Who it affects
The report targets individuals facing difficult acoustic localization tasks in cluttered spaces. It demonstrates how AI agents can build custom utility tools for specific, one-off problems involving consumer hardware.
What is still unknown
The material does not state if the code is open source or available for public download. It also does not detail the specific audio processing algorithms used beyond general triangulation concepts.
What changes, what does not
The system successfully located the source of the beep. Willison found an old carbon monoxide detector needing batteries in the identified area. The initial automatic location estimation required manual correction.
Methods and limits
The system plays noises between devices to estimate their relative positions. This automatic method performed poorly, prompting a "Rough mode" where users drag icons to set approximate locations. These rough positions are then adjusted based on sound data. The devices listen for the beep, and users can test the system by clapping. Beeps appear as X marks on a map. Users can replay audio from each device to confirm the sound matches the mystery beep. Once confirmed, the system measures beep intervals and uses triangulated locations to pinpoint the source. Willison noted the system worked as intended, providing a position within about 40cm.reported He also referenced a 2007 paper by Chunyi Peng and colleagues titled "BeepBeep," which describes a high-accuracy acoustic-based ranging system using commercial off-the-shelf devices.reported
Why it matters
- It shows AI agents can integrate multiple consumer devices into a cohesive sensing network for physical tasks.
- The need for manual "Rough mode" highlights current limitations in automatic acoustic positioning on standard hardware.
- The reference to a 2007 paper suggests AI can retrieve and apply older academic research to modern tool-building.
- It demonstrates a practical workflow for debugging environmental issues without specialized equipment.
Topics: AI agents, acoustic localization, Python, consumer hardware, Simon Willison
Updated
Then / now / next
Finding intermittent beeps required manually searching through cluttered boxes.
An AI-built tool can triangulate the source using multiple consumer devices.
The source material does not specify future developments or next steps for this tool.
Story timeline
- Then Finding intermittent beeps required manually searching through cluttered boxes.
- 09 Nov 2026 Locating the source of a mysterious beep with Claude
- Now An AI-built tool can triangulate the source using multiple consumer devices.
- Next The source material does not specify future developments or next steps for this tool.
The strongest doubt
Counterpoint from the sources
The automatic acoustic positioning between devices performed poorly, requiring significant manual intervention to establish baseline locations.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
What devices were used to test the beep locator?
Willison used two iPads, a phone, a MacBook Pro and a Surface Ultra as listener devices.
How accurate was the final location estimate?
The system provided a position within approximately 40cm of the actual source.
Did the automatic location estimation work well initially?
No, the automatic estimation did not work particularly well, requiring manual adjustment in Rough mode.
Connected stories
Reading receipt
- The system uses multiple devices to listen and estimate locations via sound.
- A manual rough-mode adjustment improved initial positioning accuracy.
- The tool located a carbon monoxide detector within 40cm of its true position.
Still open: Is the Beep Locator code publicly available for reuse?
Still open: How does performance vary in rooms with different acoustic properties?