CEO IDRobot | Built ClawBox: a 549 EUR always-on AI box that runs 24/7 on 15W. Robotics, edge AI, open source. Replacing cloud subscriptions with hardware.
Most browser automation breaks because it runs on servers with clean, predictable fingerprints. Real sessions don't look like that. Running Playwright from your own hardware — with your own cookies, network, and IP — changes what actually gets through.
Your AI assistant should do the boring work for you. Real tasks people run on ClawBox: inbox triage, drafting replies, posting to socials, reaching sites that block bots, daily reports — all on hardware you own. 🦀 clawbox.tech/go/x
67 TOPS on 7–15W. The Jetson Orin Nano isn't a Pi with a GPU bolted on — it's a purpose-built inference box that runs 7B models at real speed. Edge AI is past proof-of-concept.
The EU AI Act raises a question most cloud AI tools struggle to answer: where does data go, who processes it, and on what legal basis? Local inference answers all three by default — what never leaves the box needs no DPA.
Everyone asks what an 8GB Jetson can actually run locally. Short answer: more than you'd think — a real assistant that drives your browser, reads your inbox, and never phones home unless you let it. We wrote down the honest limits too. 🦀 clawbox.tech/go/x
ClawBox v3.1.0: your agent now backs itself up via ClawKeep, shows up natively in the iOS Files app, and responds faster. Local-first AI that ships improvements — without touching a cloud server.
Browser automation that runs 24/7 from your own hardware doesn't depend on a third-party API staying up. Playwright a local agent on a Jetson Orin = forms filled, pages monitored, sessions logged — while your laptop is off.
Raspberry Pi or Jetson Orin Nano for local AI? The Pi is cheaper; the Jetson actually runs the models at usable speed. We put both on the bench — tokens/sec, watts, setup pain. No spin. 🦀 clawbox.tech/go/x
Small ClawBox 3.1.0 detail I care about: App Store discovery makes local AI feel less like a lab project. The goal is not another dashboard; it is an always-on agent box that a normal user can find, install, update, and keep accountable.
Local-first AI changes the failure mode: when an agent fills a form, watches a page, or updates a file, the session and evidence stay on hardware you control. The useful part is not autonomy by itself; it is autonomy you can inspect after it runs.
The practical test for browser agents is not a perfect demo. It is whether the job can wake up tomorrow, use the same signed-in session, leave logs you trust, and stop before crossing a permission boundary.