Perceptron AI

The layer of intelligence that bridges the physical and digital worlds.

[Perceptron AI @perceptroninc]

Today we're releasing Mk1.5: a new intelligence layer for embodied agents.

It flies drones, controls quadrupeds, powers smart glasses, tracks objects, searches the web, reasons visually, and dispatches its own sub-agents.

One model, no platform-specific retraining. 🧵

[Perceptron AI @perceptroninc]

Everything is open-source: run your own model against EgoHome and tell us where it lands.
Dataset: https://t.co/x9qbRaQE17
Code: https://t.co/MXidXdfYDU
Post: https://t.co/2i17qf8aDP

[Perceptron AI @perceptroninc]

Introducing EgoHome: a new open benchmark for dense, sub-second egocentric understanding.

Human-annotated household-chore video with 4 annotation types on every clip: hand boxes, subtasks, hand actions, hand captions.

It features over 750 segments under 500ms.

Data, metrics, and judge prompts are all open. 🧵

[Perceptron AI @perceptroninc]

We compare Perceptron Egocentric, at its default 4 fps and an 8 fps option, with GPT-6 Astra, Gemini 3.8 Flash and Elorian’s Glimpse, each at 4 fps and 1 fps, and with Pantheon’s Argus and NVIDIA’s FLORA. The Perceptron Egocentric harness matches or beats every other system on subtask labeling.

[Perceptron AI @perceptroninc]

Read our full results: https://t.co/rsgEOE711D

[Perceptron AI @perceptroninc]

Today we’re releasing Perceptron Egocentric 1.0: The frontier for egocentric annotations, now through a self-service API.
• Fine-grained subtasks matching Astra quality at ~1/8 the cost
• State-of-the-Art 2D and 3D hand tracking
• Individual timelines for every action on each hand and robot gripper

[Perceptron AI @perceptroninc]

Perceptron Egocentric is fully self-service: send up to 1,000 videos per batch request, poll each job, and download the labels. Inspect every job through your usage dashboard.

[Perceptron AI @perceptroninc]

ReMind(Leejun Kim, Maia Le, Matthew Lam, Alesia C): A voice-led memory-game companion for dementia care, where Isaac 0.5 fine-tuned to control an SO-101 arm shuffles objects for residents to put back in order.

[Perceptron AI @perceptroninc]

We were blown away by the projects we saw at @DubHacks this weekend. Teams spent the weekend building with Mk 1.5 and Isaac 0.5 on our Perception and Robotics tracks, and they came up with some incredible applications in just 24 hours. We've linked some of our favorites below and can't wait to see what these students build next!

[Perceptron AI @perceptroninc]

Robo-View (Arda Egrioglu, Demirhan Celik, Kale Turner, Ryan Houshyari): Isaac 0.5 fine-tuned to control an SO-101 arm as a robotic camera operator, orbiting a product in three cinematic camera moves on voice or text command and returning a finished video through an app.

[Perceptron AI @perceptroninc]

Delta (Sanjay Nagesh, Subinay Kalya, Jinay Doshi, Aditya Gajula): Version control for hardware, where an overhead camera logs every change to a breadboard and Mk1.5 identifies each part added, moved or removed.

[Perceptron AI @perceptroninc]

PollenSight (@ananya_p_, @floraysm, @raikaroyc, @FalishaAmir): Turns field video of bees into a per-flower record of pollinator visits, using Mk1.5 to verify insect tracks and map flowers.

[Perceptron AI @perceptroninc]

OpenCanopy (@jstnzhou, @ccs_kuma, @BBioren): Uses Mk1.5 to audit land for EU deforestation-law compliance from free satellite imagery, sweeping every post-2020 clearing on Earth to flag forest-to-plantation conversion.