Standard Intelligence is a US computer-use-focused frontier-AI startup that emerged in 2024 and was founded by Galen Mead and Devansh Pandey. On April 30, 2026 it raised $75M at a $500M post-money valuation, led by Sequoia and Spark Capital, a figure The Information reported as a 16× markup over its late-2024 seed valuation (Source: theinformation.com). The company describes itself as a "neolab," a small team operating like a frontier lab on a much smaller capital base.
Snapshot
Valuation
| Date | Amount | Round / Event | Source | |
|---|---|---|---|---|
| 2026-04-30 | $500M post-money | $75M raise led by [[entities/sequoia-capital | Sequoia]] + Spark Capital; reported as 16× the late-2024 seed valuation | theinformation.com |
| 2024 (late) | ~$31M post-money (implied) | Seed (16× compression baseline) | (derived) |
Founders and team
Standard Intelligence was founded by Galen Mead (CEO, age 21) and Devansh Pandey (President, age 20). Both founders are early-twenties technical co-founders, described as the youngest CEO/President pairing among frontier-AI labs (Source: theinformation.com).
Andrej Karpathy joined as an advisor after Mead and Pandey ran a Toyota RAV4 driving demo on a downtown San Francisco dashboard for him; the RAV4 demo is cited as the artifact establishing the company's computer-use credibility (Source: theinformation.com).
Technology: FDM-1
Standard Intelligence is building FDM-1, a "forward dynamics model" trained with roughly 10,000× more compute than the company's prior model and primarily on videos of people using computers. The company's central technical bet is that screen-recording video of human workflows is a richer training signal for computer-use agents than text-only or screenshot-based training (Source: theinformation.com).
The "forward dynamics" framing, common in robotics, predicts the next observation given an action and a current state. Applied to computer-use, FDM-1 attempts to model what the screen will look like after a user clicks, types, or scrolls, rather than only what action a user would take next.
Competitive context
As of April 2026, the neolab category in computer-use included Anthropic's February 2026 acquisition of Vercept (a computer-use team integration), AGI Inc. as a separate computer-use entrant, and Standard Intelligence, described as the most-funded standalone player. The established hyperscaler-backed counterparts are OpenAI Operator, Anthropic Claude Computer Use (following the Vercept integration), and Google Project Mariner (Source: theinformation.com).
The Information characterized the company's 16× valuation jump over roughly 18 months on a small team as an example of a "neolab fervor" investment pattern in early 2026: mid-eight-figure rounds at nine-figure valuations for sub-30-person teams positioning against the multi-billion-dollar frontier labs. The Karpathy advisorship and the Anthropic-Vercept acquisition are cited as signals that computer-use is a recognized strategic axis even as the broader AI-coding-agent and agentic-AI markets shake out (Source: theinformation.com).
Relationships
- related: Anthropic (Vercept acquisition), Andrej Karpathy (advisor), Agentic AI, Agent Architecture Patterns.
- depends-on: Sequoia Capital (lead investor), Spark Capital.