The problem
Today's AI fails at basic visual and spatial reasoning, so CAD, robotics and engineering work stays manual
What they do
Research lab building AI models that reason directly over images, video and spatial layouts
Traction Pre-product, pre-revenue; model and API waitlist only, pilots in discussion
Hiring trend
- Open roles
- 6
- Ops roles
- 1
- Headcount
- ~26
13 staff in spring 2026 to ~26 (Tracxn, est.); plans to double or triple
Hiring in
- Palo Alto
- Remote
Latest news
Oct 2026
Launched Humanity's Sixth Sense visual-reasoning benchmark with Scale AI; first model due later 2026 ↗Funding
- Last round
- $55M
- Stage
- Seed
- Date
- Apr 2026
Led by Striker Ventures, Menlo Ventures, Altimeter (co-leads); NVIDIA, 49 Palms, Jeff Dean participating
- Hiring+16 open roles on the live job board (3–9 roles = +1)
- Funding recency+1.5Seed · $55M, Apr 2026 (6 months ago): 6 months or less = +1.5
- Traction—Pre-product, pre-revenue; model and API waitlist only, pilots in discussion
- Red flags—None on file
- Total2.0Stored score → Warm-up
Seed rule: seed-stage without strong traction starts in Warm-up whatever the score (Steady by score alone).
Hiring and funding from the stored fields add up to +2.5. The stored score of 2 includes a judgment call the fields don't capture (often a boundary month or a round we don't count), so traction and red-flag points can't be split out exactly.
Distance (Unproven) is a separate read on how long the lead can last. It doesn't change the score.
Distance
Unproven
Strong ex-DeepMind team and $300M valuation, but no model shipped yet and OpenAI and Google DeepMind ship multimodal models to the same buyers.
Sources: TechCrunch Build Mode, Jul 2026 ($300M valuation before a product) ↗ · Elorian models page (first model 'later this year', waitlist) ↗
Open ops roles · 1
Last checked Oct 25, 2026