SI Agent News
ConfirmedClaude

Claude designs and prototypes functional electronic hardware

Anthropic collaborated with Diode Computers to show how Claude can now handle complex geometric reasoning and manufacturing constraints for circuit boards.

69%Fable 5.2 score on PCB Gym benchmark

The short version

Why it matters

This shows AI moving from simple text generation to complex, multi-step engineering tasks that require spatial reasoning. It suggests a future where hardware prototyping is significantly faster by automating the tedious layout and manufacturing checks.

From schematics to prototypes

Anthropic recently collaborated with Diode Computers to test how well Claude could handle the end-to-end process of electronic hardware design. While previous models struggled to arrange components or manage physical layouts, the newer Claude Fable 5.2 model successfully designed three distinct devices. These included a speakerphone, a voice-assistant hub, and a handheld virtual pet.

The process began with simple prompts outlining hardware requirements, such as a USB-C speakerphone with Bluetooth support. Claude then proceeded to write the specifications, draw visual schematics, and choose component placements on a printed circuit board (PCB). It even designed 3D-printable enclosures and wrote the necessary firmware to make the devices functional.

The final results were prototypes that still work today after weeks of use. Diode assembled the boards and ran the firmware under Claude's direction. This marks a shift from the AI simply suggesting parts to actually directing the production of physical hardware.

Geometric reasoning and routing

A major leap for the Fable 5.2 model is its ability to reason through complex geometric problems. In earlier tests, models often placed supporting parts too far from chips or oriented connectors incorrectly. Fable 5.2, however, produced compact and practical layouts by prioritizing high-speed signals and high-current connections first.

For the voice-assistant hub, Claude managed long tracks from a 100-pin connector by keeping pairs of copper tracks the same length and correctly spaced. It even added

Self-correcting design logic

One of the most significant improvements in the newer model is its ability to revise its own work without human intervention. When the back of a circuit board became too crowded during the speakerphone design, Claude did not simply try to patch the issue with extra jumps between layers. Instead, it swapped signals on interchangeable pins and redesigned an entire corner of the board to maintain a clean layout.

This behavior shows a level of reasoning where the AI recognizes when a design path is becoming unusable. By proactively choosing to start fresh on a specific section rather than forcing a flawed solution, the model demonstrated a sophisticated understanding of spatial constraints. This ability to "rip and reroute" helped ensure the final boards were practical for manufacturing rather than just theoretically possible.

Custom tools and signal safety

To handle the complex connections of the voice-assistant hub, Claude interacted with KiCad, a common circuit board design program. While the standard router in that software draws one track at a time between two points, Claude determined that this would be too tedious for long tracks. To solve this, it wrote its own custom router algorithm that divided the board into a grid to find clear paths for each track in seconds.

Beyond just finding paths, Claude also demonstrated an understanding of signal integrity. It noticed an audio clock track running too close to the Ethernet tracks and adjusted its custom router to leave more room around those sensitive areas. This indicates the model can balance high-level automation with the specific technical requirements of electrical engineering, such as preventing noise interference between different components.

Acoustic and physical modeling

Claude also performed complex calculations to ensure the physical enclosures functioned as intended. When designing the voice-assistant hub, it calculated that a 25 mm-long microphone port would resonate near 3.4 kHz. Because this frequency falls right in the voice band, Claude identified it as a potential issue for clear audio and reasoned that a different shape was necessary for better performance.

The model proposed a wedge-shaped enclosure that was tall at the rear and stepped down at the front. This design shortened the ports to about 10 mm, which raised the resonant frequency to about 8.9 kHz at the edge of the speech range. By modeling these physical effects, Claude was able to provide a design that accounted for how sound waves interact with the device's shape.

Sources and links

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

Listen to the AI Wire headlines

Listen to the AI Wire

Listen to our daily AI shows every morning: on your commute, in the shower, wherever you are

Listen to the daily shows

First reported by anthropic.com · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

More AI news