Steve Blank, the guy who basically wrote the book on modern startup methodology, has a message for founders: if your company is more than two years old, your business plan is probably garbage. The AI wave has moved that fast. Venture capital agrees. Two-thirds of all VC dollars in 2025 went to AI deals, leaving everyone else fighting over scraps.
The technical argument is straightforward. Tools like Claude Code and OpenAI Codex have collapsed development timelines from months to days. An MVP used to prove your team could build something. Now it proves you spent an afternoon with the right prompts. And the "data moat" that startups clung to? Foundation models have ingested enough public knowledge that proprietary datasets matter less. Synthetic data and RAG architectures let competitors train solid models without years of data collection.
Blank isn't just saying "add AI to your startup." He's arguing that business assumptions expire faster now. Hacker News commenters pushed back on some of his examples, noting that faster code generation doesn't solve bottlenecks in system design, pricing, or UX decisions. Fair point. Still, the speed at which your founding assumptions go stale is real. Ask yourself: does your 2024 strategy still hold when AI agents can test multiple product variations in parallel for the same cost as one?
For those building agents, Blank highlights a shift worth watching: software is moving from interface-based to outcome-based. Users won't click through menus. They'll describe what they want, and agents will handle it. A whole different product category.