Quick answer

10 predictions for 2027: GPT-6 and Claude 5 release (probably). AI agents become reliable for narrow tasks. Open source matches frontier. EU AI Act fines hit Big Tech. AI-native browsers cross 100M users. Music industry forced to legalise AI. Universities formalise AI in curriculum. AI energy demand triggers grid expansion. ChatGPT competitors gain market share. AGI debate gets louder but no consensus.

A year from now, here are the bets we would actually take. Mix of confident and speculative — flagged in each.

High confidence (>80%)

  • GPT-6 and Claude 5 release — both labs are working on next-gen now
  • Reasoning models become default — chat-only models phased out
  • Open source matches frontier in raw capability — but Big Tech keeps integration advantage
  • EU AI Act first major fines — Big Tech will be the test cases
  • AI energy demand triggers permitting reform — grid expansion accelerates
  • AI-native browsers cross 100M users — Chrome's grip loosens

Medium confidence (50-80%)

  • AI agents become reliable for narrow workflows — full autonomy still flaky
  • Music industry forced to legalise / license AI training — RIAA settlements
  • ChatGPT loses market share to Claude and Gemini — gap narrows
  • Universities formalise AI in curriculum — required courses appear

Low confidence (speculative)

  • AGI consensus achieved — unlikely; the debate will continue
  • A major AI safety incident — possible but no specific prediction
  • AI replaces a whole job category overnight — unlikely; gradual
  • Quantum computing meaningfully impacts AI — too early

Things people predict that probably will NOT happen

  • Mass unemployment — AI displaces specific roles, not whole sectors
  • Hollywood replaced by AI — actor unions block this aggressively
  • OpenAI losing its lead — possible but unlikely in 2027
  • AGI achieved — depends on definition; no widely-accepted milestone in 2027

The honest mid-2026 take: AI progress is still steep but predictable. The discontinuous shifts of 2022-2023 are unlikely to repeat. We are in the "infrastructure" phase — applying current capabilities broadly, not waiting for the next big model.

Related reading

The State of AI — May 2026 Mid-Year Report Why China's AI Is Catching Up to America's The Open Source AI Movement — Why It Matters

Bottom line

2027 will be more "polishing 2026" than "transforming 2026". The next dramatic shift probably comes in 2028-2029 when agents finally work reliably. Until then, keep iterating with what you have.