What do you mean an AI bubble? It’s the best time in history to start a company

• New company formation has surpassed the pandemic spike: According to Stripe data, the number of newly formed corporations/businesses in 2025–2026 nearly doubled year over year, surpassing the peak seen during the pandemic.
• More substance and more speed at the same time:
  • It is no longer limited to lightly built "vibe coding" products; average revenue and success rates per business have actually risen compared with the past.
  • The time it takes to reach recurring revenue milestones such as $1 million/$10 million ARR has been cut in half compared with the SaaS boom era.
  • 20% of startup founders generate their first customer payment within 30 days (versus 8% in 2020).

② Silicon Valley concentration vs. distributed hubs

• The long tail is beginning, and capital is flowing into Silicon Valley: As barriers to entry fall, startup activity is increasing everywhere in the world.
• At the same time, there is a tendency to start businesses elsewhere, build meaningful growth there, and then ultimately gather in Silicon Valley to gain further momentum.

③ The rise of domain experts and "1–2 person AI companies"

• Innovation in spaces software never reached: Domain experts who are not traditional programmers—teachers, auto mechanics, on-the-ground operators, and others—are directly applying AI to problems and rapidly building companies worth millions to hundreds of millions of dollars. (For example, the AI tool for teachers, 'MagicSchool', and 'Medvi', a two-person company aiming for $100 million in revenue)
• The normalization of one-person / small elite teams: As development, marketing, and sales become vertically integrated inside platforms like Replit, large 2–3 person companies that manage AI agents instead of people will likely become increasingly common.

④ Human territory that remains in the AI era

• Cultural intuition and domain understanding: Because AI models operate within their training data, the ability to sense the edge of what is possible and catch the latest cultural trends (for example, looksmaxxing apps) remains uniquely human.

Creativity outside the distribution (paradigm shifts): AI will dominate polished problem-solving within given frameworks, but creativity that breaks entirely new paradigms is still likely to come from human serendipity and insight.

⑤ The future moat for software

• A structure AI labs cannot monopolize: Just as in the dot-com crash or the Microsoft/Google monopoly debates, AI labs will not capture all the value. AI will seep in as a general-purpose foundational technology, like the internet or Linux.
• Software itself is not a moat: The ability to simply build code is not a moat; traditional business moats such as liquidity/coordination, brand, and network effects will still play a decisive role.