AI 2027 Milestone Tracker

Narrative

The period from January 2025 through April 2026 has produced a rich and often contradictory body of evidence against the AI 2027 report's milestone predictions. On the 'shipped' side, the agentic AI layer that AI 2027 predicted for mid-2025 has clearly arrived: OpenAI launched Operator in January 2025, integrated it into ChatGPT as a unified agentic system ('ChatGPT Agent') by July 2025, and released the enterprise-grade 'Frontier' platform in February 2026; Anthropic's Claude Code became a widely adopted agentic coding tool; Google DeepMind shipped Gemini 3 (November 2025) with multi-agent orchestration and a reimagined Deep Research agent; and Anthropic's Model Context Protocol crossed 97 million installs by March 2026, becoming foundational infrastructure. Flagship model releases — GPT-5 (August 2025), GPT-5.1/5.2 (November–December 2025), Claude Opus 4.5 (November 2025), Gemini 3 (November 2025), Grok 3 (February 2025) — broadly match AI 2027's qualitative prediction that '2025 AIs function more like employees.' The Stanford HAI 2025 AI Index confirms that SWE-bench scores reached 71.7% by end-2024 (up from 4.4% in 2023), and METR's time-horizon metric showed the frontier doubling approximately every 7 months. OpenAI's annualized revenue hit ~$20B, slightly ahead of AI 2027's $18B prediction. On the 'unshipped or contradicted' side, the authors themselves revised their median superhuman-coder timeline from 2027–2028 to approximately 2032 by December 2025 (a 3–5 year slip), citing modelling errors in AI R&D automation assumptions. Their own grading report found overall quantitative progress at only ~65% of predicted pace: SWEBench scores reached 74.5% rather than the predicted 85% by mid-2025, no leading AI company conducted a substantially larger training run than GPT-4.5, and AI R&D software uplift remains well below the 1.9x ratio predicted for end-2026. Critically, METR's randomized controlled trial found that early-2025 AI tools actually made experienced open-source developers 19% slower, not faster — a sharp contradiction of AI 2027's productivity uplift assumptions. The safety framework story is equally complex: Anthropic activated ASL-3 safeguards (an AI 2027-adjacent milestone) in May 2025, but then dropped its hard pause commitment entirely in February 2026 under competitive and political pressure, replacing rigid guardrails with nonbinding 'public goals.' The Pentagon simultaneously threatened Anthropic with blacklisting over safety red lines, illustrating that geopolitical friction — not just technical progress — is shaping deployment trajectories in ways AI 2027 did not adequately model.