Pitch Presentation

Hippocampal

Video reasoning is the next fundamental intelligence paradigm.

Press → or click to beginSwipe to begin

Language Models are Limited

Even the most advanced LLMs fail at the simplest tasks for kids.

Video Models Start to Reason

The real substrate of spatiotemporal intelligence has arrived.

“Navigate through the maze from green start to red flag end.”

Explore →

Our model0.708

Sora 20.564

Veo 30.505

#1 HuggingFace Paper of the Month — February 2026

ICML 2026

56 authors · 32 universities · 5 continents

VBVR — A Very Big Video Reasoning Suite (arXiv 2602.20159)

Results

Our model beats both OpenAI Sora and Google Veo by more than 25%.

Our model scores 0.708 on the VBVR benchmark. OpenAI's Sora 2 scores 0.546. Google's Veo 3 scores 0.480. This isn't a cherry-picked metric — it's a systematic evaluation across 200+ reasoning tasks. We didn't just build the test. We built the model that passes it.

Our model0.708

Sora 20.546

Veo 30.48

Traction

The world sees us.

Organic coverage from major AI influencers and content creators — no paid promotion.

200+

Media pieces

1M+

Views

100K+

Downloads

Platforms

All organic, from leading AI influencers.

WeChat

YouTube

Bilibili

Little Red Note

Traditional Media

X / Twitter

TikTok

What happens when video models learn to reason

Emerged Abilities

Embodied Operating System

Video-audio model is the operating system for next generation human-machine interfaces.

Before Trained for Reasoning

After Trained for Reasoning

Language is a subtask of video reasoning.

Function tool calling leads to language model agents. We wonder what is the model context protocol for video models.

Language reasoning as a subtask of video reasoning

Robotics is a subtask of video reasoning.

It turns out inverse dynamics computations in human brain happen in visual cortex.

Visual cortex brain mapping — inverse dynamics in V1-V4 areas

In-Image Instructions

Prompting is a very unnatural way for artists to create. Direct editing on visual content is definitely the future UI of creative AI.

From research to revenue

Growth Strategy

Active Products

Monetization over Creative, Data and Eval first.

We play the Anthropic game, where monetization over chatbot is a day-one strategy.

Potato Studio

Draw on images to generate videos.

Data Fountain

Generate video reasoning datasets on demand.

Video Eval

Real-time rule-based video evaluation.

50+

Artists signed up

Potato Studio

Enterprises in conversations

Data Platform

Active lab users

Video Eval

Vertical Domains Are When The Actual Game Starts

Our continued foundational research guided by deep understanding of intelligence will find us the Claude Code moment.

Industrial domain reasoning — tile, screw, capsule, metal plate

Manufacturing & Factories

Medical domain reasoning — skin, colon, thyroid, brain

Health & Medicine

Sports Analytics

Weather & Climate

New Results

prompt >“Colorectal cancer is the third most common cancer worldwide. Please highlight polyps in the video.”

Input

Output

Open sourcing is the best marketing strategy in AI.

The Ecosystem Moat

We open-source the full stack — data, model, and infra. Smartest people build on us, with us, beyond us.

Hover a pillar to explore · click a node to open the repo

Data5 repos

Open-source procedural video data generators — anyone can create training data.

Model1 repos

Open weights on HuggingFace — the community fine-tunes and extends.

Infra3 repos

Open-source evaluation, data pipelines, and training code.

Data

Open-source procedural video data generators — anyone can create training data.

Read The Chart ↗BFS ↗Visual Jenga ↗Hit Target After Bounce ↗Shortest Path ↗

Model

Open weights on HuggingFace — the community fine-tunes and extends.

VBVR-Wan2.2 ↗

Infra

Open-source evaluation, data pipelines, and training code.

Eval ↗Data Scaling ↗Training ↗

Team

Intensity Moat

All Gen-Z. First batch of VBVR team to join fulltime. Three years building together. High-impact projects, minimal resources. We don't wait for alpha — we just ship alpha.

Unfair Advantage

100 university Discord channels. 80 WhatsApp groups. One afternoon to mobilize 500 contributors. We call it the Boba Tea strategy — and no data platform on earth can match our cost or speed.

Hokin

CMU

Mark

UC Berkeley

Sara

UCSD

Zehong

UCSD

Kaia

JHU

Market

Video reasoning is a trillion-dollar opportunity.

TAM

SAM

SOM$10B

TAM

~$3T

World reasoning intelligence

SAM

~$500B

Video reasoning — training, inference, evaluation

SOM

~$10B

Synthetic data, model APIs, and creative tools — where we play today

Markets that need video reasoning

RoboticsManufacturing & QCMedical ImagingCreative & MediaAR / Spatial ComputingWeather & ClimateSports AnalyticsAutonomous DrivingConstruction & Architecture

Investment Memo

1.Video models start to reason. This is the ground floor of a new AI category.
2.Gigantic open-source ecosystem moat.
3.Ability to compete with OpenAI Sora and Google Veo by outperforming 25%.
4.Crazy organic traction from the creative space.
5.Unique unfair advantage to mobilize 1,000 people at cost of $3/hr.
6.Gen-Z only team — Intensity is the only moat.
7.Ready-to-go monetization on selling creatives, data, and evals.
8.Further research validation on verticals over trillions dollar opportunity.

hokinxqdeng@gmail.com

1 / 18