Interview
Inside the Fastest-Growing Category in AI: Scott Wu, CEO of $26B Cognition
- Cognition projects that within three years, coding agent adoption will expand from negligible levels to near-universal usage by all software engineers and many non-engineers, with platform usage expected to grow by a factor of 11 to 12 times and code shipped volume increasing roughly 7 times over the last six months.
- Revenue and adoption growth targets are set at a rate where a 3x year-over-year increase would still lag behind the broader market, driven by the expectation of an "abundance era" occurring within 12 months where AI handles personal decisions and daily details.
- By 2031 or later, the company anticipates a shift from manual development to "single-use" and "self-driving" software generated on the fly, which will render current methods primitive and transform government services like the DMV.
- Within the next 12 months, enterprises are expected to increasingly seek third-party providers to mitigate risks associated with model obsolescence, necessitating organizational changes in planning, specs, design, and user research, while model neutrality across providers like OpenAI, Anthropic, and Google is viewed as a key long-term value proposition.
- Significant infrastructure and product integration milestones include securing a larger San Francisco office by the end of last year for combined Cognition and Windsurf teams, maintaining independence to prevent a single-entity future, and a gradual integration process rather than an immediate merger.
- The market outlook for late 2025 to early 2026 indicates a surge in interest for asynchronous agents and integrated local-cloud solutions, occurring alongside a cautious timeline where AI does not immediately render humans irrelevant despite real "recursion" capabilities.
- While intelligence problems may be solved in the long term, practical hurdles including hardware availability, processes, and amorphous tasks are expected to prevent the resolution of all societal issues within six months, though an "event horizon" for predictive difficulty is anticipated eventually.
- AI is predicted to evolve into a fundamental policy issue within the next few years, transitioning from controlled access to an open system similar to the early internet, with a specific goal to avoid a future where a single entity rules global intelligence.