Opposable lets AI tap and type on Android phones and iPhones
Opposable is a new tool that lets an AI agent control Android phones and, via a companion computer, iPhones. The app uses Android accessibility services to read the screen, sends screenshots to the GUI‑Owl 1.5 2B vision‑language model running through llama.cpp, and then issues taps, swipes and typing based on the model’s output. Tasks can be driven by a paired desktop AI agent or by an on‑device…
Key points
- Opposable uses GUI‑Owl 1.5 2B vision‑language model via llama.cpp to interpret screenshots and issue taps.
- Example tasks cost $0.44 and take 50‑96 seconds, with success on about three‑quarters of unseen tasks.
- Users must sideload an APK, grant accessibility permissions, and keep the API token secret like a password.
The blog post that announced the service shows three example runs. Changing the system theme to dark mode and setting a five‑minute timer took 50 seconds and required 17 tool calls. Looking up the Apollo 11 launch date and crew took just under two minutes and cost $0.44. Opening the clock app and switching to the stopwatch tab on a virtual Android device took about 96 seconds. The authors note that on eight tasks the model had never been tuned on, it succeeded four times in one round and three times in another, roughly a three‑quarters success rate. The write‑up also admits the system is “not particularly fast” and its accuracy is “questionable”.
Installation involves sideloading an APK and granting accessibility permissions, and the website warns that anyone holding the API token can operate the phone, likening the token to a password. The team did not respond to requests for comment, and the site does not mention any accessibility‑focused use cases.
Reach new heights of laziness by burning tokens to let an AI tap your smartphone screen for you
The Register AI/ML · 21 October 2026
Loading the full article…
This text was published by The Register AI/ML and written by Brandon Vigliarolo. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Aws introduces Adaptive agentic intelligence framework with Strands Agents · 1 src
- Walmart launches Agent Gateway for AI shopping agents · 3 src
- AI agent fixed the bug, but what else did it change? · 1 src
- Hugging Face releases TorchNeuron beta for easier Hugging Face models on AWS Trainium · 2 src
- LTSharpe and Claude AI boost Oblivion remastered performance by up to 68% · 2 src
Comments
via GitHub Discussions