Simon Willison used Claude Code to turn spare phones and tablets into a finder for a mystery beep

What you could do with it

• A homeowner could scatter spare phones and tablets around a noisy room and use the web-based triangulation tool to pinpoint where an intermittent beep is coming from, saving them from digging through stacks of unpacked boxes ¹.
• A developer or hobbyist could prompt an AI coding assistant like Claude Code to build a custom Python web app that turns everyday devices into a coordinated sound-detection network using only browser microphones and WebSockets ¹.
• Anyone could test and calibrate such a sound-location app by clapping in different spots around a room and watching whether the app accurately marks each clap on its map before trusting it for real discoveries ¹.

A mystery beep

Our home includes a storage room packed with boxes from our previous relocation that remain unpacked. Approximately two weeks prior, an intermittent beeping noise originated from that space. The intervals between beeps were lengthy enough that attempting to pinpoint the sound by remaining in the room proved exasperating, and I wished to avoid rummaging through each container. Consequently, I requested Claude Code to develop Beep Locator, a program designed to locate a beep's origin through triangulation via several devices distributed throughout the area ¹.

Building the locator

Willison began the project on his iPhone using Claude Code Cloud, running Claude Opus 5.5 in Extra thinking effort mode ¹. He told the assistant to create a Python web app that one device would run as the main server, while other phones, tablets and laptops joined through a URL and acted as listeners using their browser microphones and WebSockets — the technology that lets browsers talk to servers in real time .

During the planning stage, Claude dug up a 2007 research paper called "BeepBeep" by researchers at several universities, which described exactly the kind of software-only, speaker-and-microphone approach needed ¹. The paper showed how two-way sound sensing and sample counting could pin down distances between devices without any special hardware .

Once most of the system was written, Willison moved to his laptop — just as Fable 5.2, another Anthropic model released that month, became available ¹. He kept Claude Code running live so it could watch what was happening and make changes on the fly while he tested things in the actual room .

How it works

The app plays test chirps between devices first, letting the system estimate where each listener sits in the room. Willison also added a "Rough mode" where users can drag device icons around on the map if the automatic placement is off ¹. Once set up, every device records a rolling 60-second audio buffer and runs a simple detection algorithm based on Fast Fourier Transform — a math trick that breaks sounds into their component pitches — to spot the beep pattern .

When a beep is detected, the server plays a short correction chirp and asks every listener to upload a clip from about 0.6 seconds before to 0.8 seconds after that chirp, which lets the system line up the recordings and calculate where the original sound came from ¹. Willison also built a clapping mode so he could test the setup by making noise in different spots, marking each clap on the map to confirm the tool was working .

Finding the culprit

The system worked. It narrowed the beep's position down to within about 40 centimetres — roughly the width of a standard ruler — and Willison dug through only that specific pile of boxes ¹. Inside, it discovered an obsolete carbon monoxide alarm that urgently required fresh batteries . He was able to mark the confirmed beep in the app and let it lock onto the same frequency for future alerts .

This story reports on a trend for information only. It is not advice or encouragement to do any of it.

How good is it

The system was tested by its builder, Simon Willison, who found it located beeps to within about 40cm in his own room before he tracked down an old carbon monoxide detector ¹. That result comes only from a single personal home test, not from an independent lab or a comparison against rival tools . No other published benchmarks or head-to-head scores against competing approaches are reported in the source .

Corrections and updates

• Corrected on 10 November 2026. An earlier version called Simon Willison a 'hobbyist,' but he published the post on his well-known technology blog and is widely recognized as a software developer; the later version corrected this to 'developer.'

Key facts

• The system locates a beep to within about 40cm using multiple listening devices
• It was built by Simon Willison using Claude Code with Claude Opus 5.5, then continued on Fable 5.2
• Any phones, tablets or laptops with a browser can join as listeners via a shared URL
• Based on a 2007 academic paper called 'BeepBeep' about software-only acoustic ranging
• Found an old carbon monoxide detector whose batteries needed replacing

Sources

1. Locating the source of a mysterious beep with Claude — simonwillison.net