Edge

Last signal: Feb 13 — Karpathy's 'bacterial code' = code becomes model-agnostic bu…

Frontier models converge in capability. The moat moves up-stack to whoever owns the workflow, the data, and the user relationship. OpenAI, Anthropic, and Google compete on price, not performance. Winners build irreplaceable applications ON TOP of models.

Last signal: Jan 15 — McKinsey/BCG reports consistently showing pilot-to-productio…

Consumer AI hit mass adoption in 2023-24. Enterprise is still in pilot phase. The next 2-3 years will see massive enterprise adoption — not chatbots, but AI-native workflows replacing entire business processes. First movers get compounding advantages.

Last signal: Feb 14 — Simile raising $100M to simulate human behavior at scale

The SaaS model (human uses software tool) gets replaced by agentic systems (AI uses tools on behalf of humans). Instead of 15 subscriptions, you have agents that DO the work across systems. This kills point-solution SaaS and creates value for platforms agents plug into.

Last signal: Feb 11 — Matt Shumer: "AI can now do all of my technical work"

AI collapses the cost of building and operating a business. What required 50 people can be done by 5. What required $5M seed can bootstrap to $1M ARR. This creates an explosion of small, profitable, AI-native businesses.

Last signal: Feb 9 — Benaich's State of AI: 'embodied AI hitting its stride'

2023-25 was AI for text/code/images. 2026-28 is AI for the physical world — robotics, autonomous vehicles, manufacturing, logistics, healthcare. Companies bridging AI into physical operations capture the biggest markets.