CONFIRMED
Keybytes is a modular keyboard that starts from a single magnetic key module, and each module carries its own RISC-V microcontroller. Tom’s Hardware reports that the developer has published the project on GitHub, and Adafruit’s blog says the whole thing is MIT licensed. It is an early prototype, not a product, but it is already a promising platform for hacking.
What we know
The system has two parts. Tom’s Hardware describes a hub called the Keyword that connects magnetically to keys called Keybits. Each Keybit has its own microcontroller, so the user decides how many keys there are and how they are laid out.
According to the GitHub repo and Adafruit’s blog, each key’s microcontroller is a CH32V003, a small RISC-V part. Some secondary write-ups have said the chip model is unconfirmed. That is wrong, because the chip is named in the repo itself.
Tom’s Hardware breaks the Keyword hub into two boards. The top board holds an Arm Cortex-M4-powered microcontroller, a Bluetooth antenna and flash memory. The lower board handles USB-C, charging and a buck-boost converter. That means the design has one Arm brain in the hub and a RISC-V chip under every key.
Adafruit’s blog, written by Anne Barela on October 26, 2026, covers the mechanical and design side. Both board types are 1u, which is 19.05mm x 19.05mm, the standard keycap pitch. All the electronics were designed in KiCad. The cases were drawn in Onshape, and their STEP files are in the repo’s cad folder. Under the MIT license you can fork the boards, modify the cases and build on the work.
On timing, Tom’s Hardware cites the project journal as saying Keybytes began in March 2026. The developer posted it to Show HN in late October 2026. Adafruit and The Register covered it on October 26, and Tom’s Hardware followed on October 27. All three describe it as an early-stage, open-source prototype.
What we don’t know yet
Firmware is the biggest gap. On Show HN, the developer said firmware is still in progress and has not been pushed to the repo. Some secondary write-ups say firmware and toolchain details are unknown, and on that point they are correct. Nothing about the code running on the CH32V003 or the Cortex-M4 has been published.
The developer also described the planned design in that thread. Treat it as forum commentary, not a confirmed spec. He said each key is meant to remember its “parent” neighbor, the one closest to the main board, and forward events toward the main board. He also said four I2C buses help reduce parasitic capacitance and let software work out where each key sits. That is an interesting approach to discovering the topology of a freeform layout, but it is a stated plan and nothing has shipped yet.
Scale is unproven. Tom’s Hardware notes that the GitHub page shows no more than two Keybits attached to a Keyword. A full keyboard built from these modules has not been demonstrated there.
There is no ship date, kit or price in any of the coverage. Tom’s Hardware only warns that the build “might end up costly.” With a microcontroller and its own board in every key, that cost is easy to see.
What it means
For makers, putting an MCU under every switch is the interesting part. Each Keybit is a small programmable node. In principle that opens the door to per-key logic, odd layouts and experiments that a normal matrix keyboard cannot support. The open KiCad files, STEP cases and MIT license mean you are not waiting on anyone’s permission to try.
For now, you cannot build a working keyboard from the repo alone, because the firmware is not there. If you want to start, the realistic step is to study the board files and the CH32V003 datasheet, then follow the repo for the firmware push. Before ordering a full layout’s worth of PCBs, check the cost per key against Tom’s Hardware’s warning that the build could get expensive.
Follow every development in one place on our board release tracker.
