Image via Unknown
Tuesday, September 9, 2025
Cognition hits $10.2B as AI coding agents go mainstream
The valuation race is heating up (yikes)—Cognition just hit $10.2B on the back of enterprise AI coding agents gaining real traction, while SharkNinja's crushing Q2 earnings shows AI-powered consumer products are moving units fast. Meanwhile, engineers are getting smart about efficiency: vector embeddings and graph databases are making AI coding agents 90% more token-efficient, which means the real leverage isn't in scaling compute anymore, it's in smarter architecture. On the policy front, Anthropic's backing California's AI safety law with a disclosure-focused approach rather than restrictions—a pragmatic move that echoes Donella Meadows' insight that real system change targets paradigms, not parameters. Here's the question: are you building toward token efficiency or raw scale?
Image via Unknown
Top Stories
SharkNinja Investor Relations
SharkNinja beat expectations with strong Q2 growth and raised full-year guidance, driven by innovative products, international expansion, and improved margins despite tariff pressures, positioning the company as a leading innovator in consumer appliances.
AI coding agents waste excessive tokens due to inefficient codebase search and context degradation; combining vector embeddings and graph databases can reduce token consumption by 90%+ while providing faster, more complete results.
Cognition
Cognition's $10.2B funding round validates the rapid maturation of AI software engineering, with the company's Devin agent and Windsurf IDE combination already generating $100M+ ARR and capturing major enterprise adoption across finance and tech sectors.
Donella Meadows Institute
Meadows identifies that effective system change requires targeting paradigms and goals rather than parameters, as complex systems are fundamentally counterintuitive and resist surface-level interventions while responding to deep structural shifts in beliefs and incentives.
Anthropic
Anthropic backs California's SB 53, which mandates transparency and safety reporting from frontier AI developers, establishing baseline disclosure requirements that prevent competitive pressure from undermining AI safety practices.
Keep Reading
Industry Voices
Boaz Barak
Researcher and Professor at OpenAI and Harvard
Bridges theoretical computer science and AI safety research, translating complex mathematical concepts into practical AGI alignment strategies.
Ashish Kumar
Research Scientist at Meta
Pushes robotics and embodied AI forward at Meta, showing how language models can control physical agents in the real world.
Scott Wu
Cofounder and CEO at Cognition
Built Devin, the first AI software engineer that actually ships code autonomously, not just generates snippets.
Greg Brockman
cofounder and president at OpenAI
Architect behind OpenAI's API infrastructure and the technical decisions that made GPT models accessible to millions of developers.
Amer Sinha
Software Engineer at Google Research
Works on Google's core ML infrastructure that determines how efficiently models train at massive scale.
Enjoyed this issue?
Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.