
Ask Wikipedia in airplane mode. An offline AI research assistant for Android made by @0xPD33
Joined November 2026
- Tweets25
- Following1
- Followers63
- Likes5
Pinned Tweet
Commonplace: offline AI research for Android, with cited sources.
Site: askcommonplace.app
Code: github.com/0xPD33/commonplac…
Community token (Base): not part of the app, not financial advice. I get a share of its fees.
CA: 0x334dbefdedc6ff68d5a883e5ea27a37a137d5ba3
The next release, Commonplace 0.4.0, gets Gemma's first word 31% sooner on a Pixel 10, fully offline: 1.86 s → 1.28 s (median, 34 questions).
The trick: the CPU and the Tensor NPU no longer take turns. 🧵
Every change in 0.4.0 passes our eval first: the same questions, graded against Claude with web search.
No cloud at any step. Your question never leaves the phone.
Try Commonplace today: github.com/0xPD33/commonplac…
Every model on Commonplace has to earn its place in our eval.
LFM3 and Spark score well on agent benchmarks. Our task is cited answers from a library, and that needs a different skill.
LFM3 got Liquid's official format and sampling. It and Spark score well on agent benchmarks, but our task is cited answers from a library.
The long-term plan: fine-tune our own small models on exactly this task. The same eval will decide whether they ship.
Keep an eye out on the next release and try Commonplace now: github.com/0xPD33/commonplac…
Same small model, same 20 questions about events in 2025 and 2026.
Without the library: 0 right.
With Commonplace's offline library: 19 right.
That's from the new benchmark. Here's what I'm building and testing next, plus a few ideas 🧵
One idea for the community token, $COMMONPLACE: reward people who test the app on their own phone, send in device benchmarks, or report bugs on GitHub.
The app still never needs the token or a wallet.
Which of these would you use first?
And if your phone isn't a Pixel, a benchmark from it would help a lot. The APK is here: github.com/0xPD33/commonplac…
One feature I'm proud of is the answer card which appears almost instantly after a prompt is sent without waiting on the LLM.
Here is a small video that demonstrates the impact.
Interested in how Commonplace works and what makes it stand out? There is a lot of documentation on the backend on GitHub (just follow the README): github.com/0xPD33/commonplac…







