Google open-sources ARTEMIS, an on-device Android GUI agent at 3s per step
aigclink · x · 2026-09-22
Google open-sourced ARTEMIS, a GUI agent that operates real Android devices: it reads the screen, finds controls, taps and swipes across apps, with logging, screenshots, replay and reports.
- Element finding: combines accessibility hierarchy/XML UI trees, OCR and vision models for Canvas/Compose/Flutter UIs, falling back to visual/coordinate taps.
- Long tasks: compresses old screenshots into summaries to avoid context blowup.
- Two modes: Flash (3–5s/step, fast observe-act loop) and Pro (15–40s/step, planning, handles 100+ step tasks).
- MCP-native: plugs into Claude Code, Codex, etc., letting AI drive real-device debugging.
A concrete step forward for mobile testing, debugging and automation.
More from coding & agent
- 53 harness bugs: an engineering team's field notes on building eval harnesses — DrDatta_AIIMS · 2026-09-22
- atuin 18.23 ships infinite searchable scrollback — shell history now keeps command output — aronchick · 2026-09-22
- Open-source coding-agent skills for Google ADK on Google Cloud, distilled from real engineering lessons — Difficult_Design6676 · 2026-09-22
- After trimming his Claude Code harness, a Max 20x user now has 9 spare hours of quota a week — carlito_17 · 2026-09-22
- OpenAI shows how a surgeon uses Codex to build tools and review PubMed literature — OpenAIDevs · 2026-09-22
- Similarity Isn't Relevance: Four Layers Every Personal Agent Memory System Must Separate — sujingshen · 2026-09-22