Developer Runs Linux and AI Chatbot on Kindle
Pragyaan Gaur's Inkwell and Quill projects show a 2015 Kindle Paperwhite running a Linux userspace and a small language model entirely on-device.
Updated on Nov. 7, 2026 in Artificial Intelligence
Live Poll
Would you try running an AI chatbot on an old e-reader you own?
Developer Pragyaan Gaur released Inkwell, an Alpine Linux userspace running on a jailbroken 2015 Kindle Paperwhite. A companion chatbot called Quill runs SmolLM2 through llama.cpp on the device's own processor, with nothing leaving the device.
Why it matters
The projects demonstrate that small language models can run locally on very modest, low-power hardware without a network connection. For readers, it highlights how older consumer devices can be repurposed for offline AI use.
Quill's quick model writes about 1.7 words per second, while its smart mode writes about 0.7 words per second. The figures come from the developer's own reported testing, and the source does not describe the methodology or output quality.
The players
Pragyaan Gaur
Pragyaan Gaur is the independent developer behind the Inkwell and Quill projects, which repurpose a jailbroken Kindle for Linux and local AI use.
The details
Inkwell is built as an Alpine Linux userspace for the e-ink device and includes a Terminal, a chess game, and DOOM. Quill writes its answer word by word, and the project is hosted on GitHub.
Timeline
The Kindle Paperwhite the project runs on was released in 2015.
The article describing Inkwell and Quill was published on 2026-11-07.
The Tech Race
The project extends the working life of the 2015 Kindle Paperwhite by turning it into a Linux computer capable of running a small language model offline. It points toward a wider push to run compact AI models on low-power consumer hardware rather than in remote data centers.
Owners of an older Kindle Paperwhite could in principle install Inkwell and run Quill offline, gaining a Linux terminal and a text-generating chatbot on a device they already own. Running the model is slow, at roughly 0.7 to 1.7 words per second, so it suits short queries rather than long conversations.
The takeaway
Older consumer hardware can often be given new life through open-source software, and compact language models make some offline AI use possible on modest devices. Readers curious about this should expect slow output speeds and experimental setup steps.
Further reading
For more on this topic, see our Artificial Intelligence coverage.
More information
Find the source code and setup instructions in the Inkwell project repository on GitHub.
Source note: This article includes information reported by XDA-Developers.
Live Poll
Would you try running an AI chatbot on an old e-reader you own?
