/AI20h ago

Instawork launches Instacore, a wearable camera rig for gig workers to gather VLA robotics training data

The four-pound, fanless rig uses five synchronized cameras.

13126139618.7K
Ryan Hickman
Chris Paxton#737
Original postChris Paxton#737
Ryan Hickman@ryanmhickman

Everyone in Embodied AI is talking about Vision-Language-Action (VLA) models. Almost no one is talking about the physical nightmare of collecting the data to train them. You can't scrape a kitchen table or a warehouse shelf from a web browser.

To get to millions of hours of diverse, real-world manipulation data, you need hundreds of rigs. But you can't buy them. If you build them out of off-the-shelf developer kits, they weigh 15 pounds, overheat in an hour, require bulky cabling, and break the first time an operator wears them on a job site.

At the Instawork Robotics Lab (IRL), we had to build our own. Meet the Instacore: a rugged, 4lb wearable egocentric data-capture engine designed specifically to survive a full shift on a standard power pack.

We didn't build a flashy humanoid. We did hard, blue-collar systems engineering:

💾 THE COMPUTE — A custom MediaTek Genio carrier board that runs completely fanless, routing 5 camera streams directly to on-board storage.

⚡ THE I/O PIPELINE — We ditched USB for industrial GMSL. Thin, ultra-flexible coaxial lines route high-speed data down to the backpack and Power-over-Coax (PoC) back up, completely eliminating batteries on the wrist.

⏱️ UNIFIED SENSOR CLOCK — The MediaTek Genio SoC drives a shared master clock straight to the ISPs driving our 5 global shutter sensors, stamping metadata at the microsecond of capture to ensure zero-drift temporal alignment.

👁️ OPTIMIZED OPTICS — 95 DFOV lenses on flexible PCB ribbon modules keep the wrist cams flat to prevent snags. A 145-degree chest camera captures the macro workspace, while a 50mm baseline rectilinear stereo head pair preserves close-up 3D mapping.

To build hundreds of these, we took over a warehouse in Mountain View in April, called it the Instalab, and brought in talented Pros with assembly backgrounds from Tesla, Apple, and Meta.

To test the systems, we had our Pros wear active rigs while assembling more rigs. The video below shows the raw, time-synchronized Foxglove playback of that exact loop.

Now, we’re shipping these units globally to scale data collection for real Pros on the job.

9:09 AM · Jun 9, 2026 · 12.6K Views
Sentiment

Positive users praise the Instacore wearable for advancing robotics in trades while negative users object to gig workers training robots as making themselves replaceable and highlight unaddressed bottlenecks.

Pos
33.3%
Neg
66.7%
6 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS1.2K
Chris Paxton@chris_j_paxton

A cool wearable data collection rig - for scaling high quality data

Ryan Hickman@ryanmhickman

Everyone in Embodied AI is talking about Vision-Language-Action (VLA) models. Almost no one is talking about the physical nightmare of collecting the data to train them. You can't scrape a kitchen table or a warehouse shelf from a web browser.

To get to millions of hours of diverse, real-world manipulation data, you need hundreds of rigs. But you can't buy them. If you build them out of off-the-shelf developer kits, they weigh 15 pounds, overheat in an hour, require bulky cabling, and break the first time an operator wears them on a job site.

At the Instawork Robotics Lab (IRL), we had to build our own. Meet the Instacore: a rugged, 4lb wearable egocentric data-capture engine designed specifically to survive a full shift on a standard power pack.

We didn't build a flashy humanoid. We did hard, blue-collar systems engineering:

💾 THE COMPUTE — A custom MediaTek Genio carrier board that runs completely fanless, routing 5 camera streams directly to on-board storage.

⚡ THE I/O PIPELINE — We ditched USB for industrial GMSL. Thin, ultra-flexible coaxial lines route high-speed data down to the backpack and Power-over-Coax (PoC) back up, completely eliminating batteries on the wrist.

⏱️ UNIFIED SENSOR CLOCK — The MediaTek Genio SoC drives a shared master clock straight to the ISPs driving our 5 global shutter sensors, stamping metadata at the microsecond of capture to ensure zero-drift temporal alignment.

👁️ OPTIMIZED OPTICS — 95 DFOV lenses on flexible PCB ribbon modules keep the wrist cams flat to prevent snags. A 145-degree chest camera captures the macro workspace, while a 50mm baseline rectilinear stereo head pair preserves close-up 3D mapping.

To build hundreds of these, we took over a warehouse in Mountain View in April, called it the Instalab, and brought in talented Pros with assembly backgrounds from Tesla, Apple, and Meta.

To test the systems, we had our Pros wear active rigs while assembling more rigs. The video below shows the raw, time-synchronized Foxglove playback of that exact loop.

Now, we’re shipping these units globally to scale data collection for real Pros on the job.

2hViews 1.2KLikes 3Bookmarks 5
BOOKMARKS8LIKES14RETWEETS1
Ryan Hickman@ryanmhickman

Precise spatial data requires perfect calibration. We built automated calibration stations using robotic arms to sweep camera targets, resolving precise camera matrices, IMUs, and wrist-to-tag spatial transforms.

We even verify sub-millisecond temporal alignment by pointing our camera nodes directly at high-speed timer monitors to guarantee perfect synchronization. 👇

14hViews 1.1KLikes 14Bookmarks 8
REPLIES2
Ryan Hickman@ryanmhickman

To top off a wild launch week, Business Insider did a deep dive today on how we're using this system to turn real-world gig work into the foundation for physical AI models:

https://www.businessinsider.com/instawork-instacore-gig-workers-wearable-camera-train-robots-data-2026-6

The physical data chasm is massive, but we are officially scaling. If you’re building VLA models and want to talk data capture, custom sensor pipelines, or calibration—my DMs are open. Let’s build. 👇

14hViews 268Likes 5Bookmarks 1
Ryan Hickman@ryanmhickman

We celebrated the launch last night in SF. No boring slides—just active servers wearing the systems in the wild while pouring drinks to prove real-world wearability, with raw data safely locked on the cards ready for ingest. We are officially in the dirt.

14hViews 325Likes 3Bookmarks 1
Ryan Hickman@ryanmhickman

In April, the Instalab was an empty warehouse floor. Within weeks, our assembly Pros and IRL Captains transformed it into an agile hardware and computer vision line, constantly rearranging the space to adapt to our calibration and QA flows.

The physical data chasm is massive, but the engine to cross it is officially running.

14hViews 476Likes 4
Zeeshan Zia@ZeeshanZiaML

@ryanmhickman Can you capture the tactile signals of juggling something in one's hand? Say as I try to bring a screw driver in the right pose after I pick it up.

12hViews 188Likes 2
Ivan Jaramillo Larrarte@pkarcdev

@ryanmhickman Very interesting work, the use of coaxial looks promising and comfortable. How can I access one unit/kit. We do have data-collection ops in Mexico, really interested to try this.

10hViews 53Likes 1
James Erwin@JTEIII

@ryanmhickman This is so important for bringing robotics to the trades. The future of construction will be totally different than today

11hViews 51Likes 1
Substance@substance_corp

@ryanmhickman The data bottleneck is bigger than human trajectories. We're building proprietary warehouse and facility datasets at Substance because robots need context about where they're operating, not just what they're doing by the way of insuring them

10hViews 35Likes 1
Ryan Hickman@ryanmhickman

@ZeeshanZiaML We are talking with some partners with tactile sensing gloves and grippers as accessories to Instacore

9hViews 64
Ryan Hickman@ryanmhickman

@pkarcdev Awesome, we'd love to talk about data ops in Mexico. DM your contact info.

9hViews 50
Stanislav Gryshyn@stas_gryshyn

@BusinessInsider getting paid to make yourself replaceable is a new kind of job description

20hViews 16
Glen Arrowsmith@garrows

@ryanmhickman Awesome! Congratulations on the launch. Very happy for you.

8hViews 4Likes 1
智享@CycleDecoded

@ryanmhickman So you're fighting the real robot war out there? Building a tool, not just talking about it. Respect. Can it survive a dropped coffee mug though?

3hViews 12
Blissy@BlissyOnX

@chris_j_paxton the fact that you cant just scrape a kitchen table is the bottleneck nobody wants to address

2h