Skip to content
Bud's Wiki

Mighty: a plug-and-play data-collection device exposing itself as ethernet + web UI

TL;DR

Section titled “TL;DR”

Asad Memon demos “Mighty” — a portable device that, when plugged into any phone or PC, advertises itself as an ethernet device with a web UI baked in. No companion app, no internet connection required. The framing in the tweet is that this zero-setup UX is what lets the author take it anywhere and collect diverse datasets on the move. Posted as a short video clip (1:02) showing the workflow.

Key claims

Section titled “Key claims”
  • Mighty exposes itself as an ethernet device with a built-in web UI when connected to a phone or PC, so no app install or internet is needed [tweet body].
  • This zero-setup property is positioned as the enabler for collecting diverse datasets in the field [tweet body].

Method

Section titled “Method”

Not described beyond what’s in the tweet text. The implementation suggestion is the standard “USB device emulating an ethernet adapter (CDC-NCM / RNDIS-style)” pattern, which lets the host see Mighty as a network peer and reach it via a browser at a known address. The web UI presumably exposes whatever capture / control surface the device needs. The accompanying video (which this page does not transcribe) likely shows phone-side and PC-side first-runs.

Results

Section titled “Results”

No quantitative results — product/demo announcement only. The tweet reports 123.8K views as of filing.

Why it’s interesting

Section titled “Why it’s interesting”

The most useful thing here is the UX pattern: any data-collection hardware that wants to be carried into the wild has to solve the “what runs on the host?” problem, and “be an ethernet device with a web UI” is one of the cleanest answers — it sidesteps app stores, OS permission dialogs, driver installs, and connectivity assumptions. That makes it a natural complement to the wearable-capture work the wiki has been tracking: SLAM on Robocap via cuVSLAM, packaged with rattler-build (Claude Code build report) discusses Robocap (a wearable multi-camera rig in the rerun-io ecosystem) running cuVSLAM, where the hard part is also data-pipeline ergonomics; and Rerun — The Data Layer for Physical AI is the broader bet that physical-AI data infra needs to be designed for “log once, replay everywhere.” Mighty is the front-of-pipeline analogue: zero-friction capture so the dataset actually exists in the first place.

See also

Section titled “See also”