Reval Labs claims Opposable lets an AI agent operate Android phones through an accessibility-enabled app and iPhones through a Bluetooth HID dongle, potentially extending computer-use workflows from desktop applications to personal mobile devices.
state: seedheat: lowuncertainty: mediumconvergesscott: lowmobile-agent computer-use smartphone-automationReval LabsOpposable
What is this?
The case centers on a Show HN launch for 'Opposable' by Reval Labs, claiming an AI agent can control Android via an accessibility-enabled app and iPhone via a Bluetooth HID dongle — extending desktop 'computer-use' workflows to personal mobile devices. The web search did not return the Show HN post itself or any direct coverage of Opposable; results show adjacent Android automation projects (AgentBlue, droidclaw, Revyl) and a different 'Reval AI Agent Platform' (revalsys.com) focused on business data agents. The Android accessibility-app approach is evidenced in the ecosystem; the iPhone Bluetooth HID dongle claim is not substantiated in these snippets.
Why it matters to Scott
Reval Labs' Opposable launch independently implements the exact mobile-agent control-path architecture Scott's canon identifies as the central technical question: accessibility-tree on Android vs Bluetooth HID on iPhone, extending desktop computer-use loops to smartphone OSes. His Agent Hands and Eyes, Agent-Native Computing, and Agent Addressability frameworks all treat this control-path split as the decisive architectural fork for V4 agent-addressable mobile devices. A shipping product claiming both paths is a concrete data point on a question his frameworks raise but do not answer.
ip:concept.agent-hands-and-eyesip:framework.agent-native-computingip:framework.agent-addressabilityip:framework.long-running-agentsip:source.give-the-agent-a-workshop-ebookradar:hsandhu-ios-on-device-agentradar:sidetap-iphone-mcp-windowsradar:mobile-harness-cross-platform-controlradar:androidlife-sustained-phone-agent-runradar:aiope-android-agent-runtimeradar:agent-desktop-accessibility-automationradar:uia-reader-semantic-desktop-perceptionradar:winmind-uia-windows-mcp
queries asked of Scott's wikis
- mobile-agent architecture: accessibility-tree vs HID vs ADB control paths
- computer-use pattern: extending desktop agent loops (perceive/reason/act) to smartphone OSes
- on-device vs cloud inference for mobile agents: local model constraints, latency, privacy
- agent memory / persistent context across mobile app boundaries
- Bluetooth HID as iPhone control vector: feasibility, permissions, Apple ecosystem constraints
- smartphone as always-on agent node: power, connectivity, background execution limits
Measured heat
now 0 pts/hpeak 5 pts/hcomments 0/hpeers p26momentum: steady2 platformsage 123h
points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion
How the heat travelled
pace: p33 vs 1247 stories at the 96h mark (now 123h old) — ahead of 3jsbench-llm-3d-generation-benchmark (1.5x), behind agent-memory-add-search-evaluation (0.8x)
Evidence (3) — ⭐ canonical anchor
Interpretation history
2026-10-11T12:22:07Z
New evidence is a first-party blog post from the creators (hn.story.50041799) — not independent validation. Engagement remains negligible across both HN posts (3 pts/0 comments on launch, 1 pt/0 comments on blog). Measured heat confirms cold: 0.17 pts/hr, 0 comments/hr, 30th percentile. Scott's explicit irrelevance vote stands. No third-party coverage, implementations, or adoption signals have emerged.
2026-10-11T11:36:50Z
evidence attached: hn.story.50041799 — First-party blog post from Opposable creators explaining their phone-agent thesis; the case is seeded and this is the primary artifact.
2026-10-08T16:13:28Z
Scott explicitly voted this case irrelevant despite its theoretical alignment with his Agent Hands/Eyes and Agent Addressability frameworks. The Opposable product ships both control paths (Android accessibility service + iPhone Bluetooth HID dongle) and the Android APK is downloadable, but the HN launch gained minimal traction (3 points, 0 comments after 51h) and no independent validation or adoption signals have appeared. Scott's downvote overrides the framework-pattern match.
2026-10-08T11:46:00Z
origin walked (opencode/cheap-glm, conf 0.85): anchor hn.story.50003787 -> echo.blog.0063135d4b by Erik (Reval Labs)
2026-10-08T11:31:10Z
grounded: converges/high — Reval Labs' Opposable launch independently implements the exact mobile-agent control-path architecture Scott's canon identifies as the central technical questio
2026-10-08T11:21:20Z
case created — The Show HN launch and downloadable Android client establish a concrete mobile-control artifact, though iPhone operation remains lightly evidenced.
Decision trace
- 10-11 23:22repriceNew evidence is a first-party blog post from the creators (hn.story.50041799) — not independent validation. Engagement remains negligible across both HN posts (3 pts/0 comments on launch, 1 pt/0 comme
- 10-11 22:42attention_routeThe editor compared this story and chose to keep watching.
- 10-11 22:36attention_candidateattach
- 10-11 22:36attachFirst-party blog post from Opposable creators explaining their phone-agent thesis; the case is seeded and this is the primary artifact.
- 10-11 22:36propose_attachFirst-party blog post from Opposable creators explaining their phone-agent thesis; the case is seeded and this is the primary artifact.
- 10-09 03:13repriceScott explicitly voted this case irrelevant despite its theoretical alignment with his Agent Hands/Eyes and Agent Addressability frameworks. The Opposable product ships both control paths (Android acc
- 10-09 02:18feedback_irrelevantScott vote via UI
- 10-08 23:27attention_communicatedReval Labs' Show HN release: Android app uses accessibility service to read screen and tap/swipe/type, exposed as MCP tools for Claude Code, Codex, Cursor. iPhone version uses a $12 USB Bluetooth
- 10-08 23:27attention_routeConcrete product shipping both control paths Scott's canon treats as the decisive question for V4 agent-addressable mobile devices. Independent validation of framework's architectural fork.
- 10-08 22:53attention_routeConcrete product shipping both control paths Scott's canon treats as the decisive question for V4 agent-addressable mobile devices. Independent validation of framework's architectural fork.
- 10-08 22:46attention_candidatecreate
- 10-08 22:46promote_anchororigin walk conf 0.85
- 10-08 22:31groundReval Labs' Opposable launch independently implements the exact mobile-agent control-path architecture Scott's canon identifies as the central technical question: accessibility-tree on Andro
- 10-08 22:21createThe Show HN launch and downloadable Android client establish a concrete mobile-control artifact, though iPhone operation remains lightly evidenced.