Simon Willison built the Beep Locator system with Claude Code to find the source of a beeping sound
Using Claude Code, Simon Willison programmed the Beep Locator tool, which uses multiple devices in a room to triangulate the source of a beeping sound. He located the source with an accuracy of up to 40 cm – it turned out to be a dead carbon monoxide detector.
Simon Willison described how he had Claude Code build him a system called Beep Locator, which solves a practical problem – a device repeatedly beeping, hidden among unpacked boxes in a room. The system runs as a Python web application, to which any number of devices (phones, tablets, laptops) can connect via QR code or URL, acting as listeners via WebAudio and WebSockets. The devices play sounds to each other to estimate their mutual position in space, or the position can be set manually in "Rough mode" by dragging icons on a map.
After capturing a suspicious beep, the system measures the interval between individual beeps and, based on triangulation from multiple devices, determines the most likely location of the sound source. According to the author, this allowed him to locate the source of the beeping with an accuracy of approximately 40 cm – after searching through the pile of boxes in question, it turned out to be an old carbon monoxide detector with a dead battery.
During development, Claude Code itself found and used the 2007 scientific paper BeepBeep (authors Chunyi Peng, Guobin Shen, Yongguang Zhang, Yanlin Li, Kun Tan), describing an accurate acoustic method for measuring distance between common devices without the need for pre-prepared infrastructure. Willison started the project in Claude Code Cloud on an iPhone with the Opus 5.5 model in Extra thinking effort mode, and then continued on a laptop, where he also tried out the newly released Fable 5.2 model. The system also includes a mechanism for remotely uploading new code to the listening devices at runtime, which the agent designed and implemented itself.
Why it matters
The case shows that agentic coding tools like Claude Code can independently find relevant academic literature and, based on it, design and implement a functional distributed system addressing a real-world problem, without prior specialized knowledge in the field of acoustic localization.
Two audiences, two different impacts
What this means
For individuals
For ordinary technically skilled users, this is an illustration of how Claude Code can be used even for unusual personal problems outside typical programming tasks – it's enough to describe the problem and let the tool design and implement a solution, including researching the necessary theory.
For a business
The example demonstrates the ability of agentic models (Opus 5.5, Fable 5.2) to independently find and apply an academic publication when designing a non-standard system, which is relevant for companies considering deploying Claude Code on more complex development tasks that also require a research component.
DevelopmentCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.