About This Tracker
Why This Exists
When AI 2027 was published in April 2025, it was the most detailed, concrete attempt to predict how artificial intelligence would transform the world over the next few years. Unlike vague forecasts, it put specific numbers on specific timelines — and invited people to hold it accountable.
This tracker does exactly that.
I built it because I believe we're at a genuine inflection point. I've spent 20+ years in tech — I was employee #20 at XING and watched social networks reshape how people connect. I founded and exited a real estate marketplace. I've led innovation labs and venture studios. I saw up close how several waves of exponential improvements were hitting our economy and our society, with ever increasing impact and disruption.
But for a while now, I've believed that AI is different — and not just in degree. The internet changed how we share information. Social media changed how we connect. AI changes the nature of intelligence itself. Every previous technology wave needed humans to build the next one. AI is the first technology that could build its own successor. That's not a gradual shift — it's a phase transition. And if the AI 2027 scenario is anywhere near reality, we may be living through one of the most important moments in human history.
We're here to document it as it unfolds.
What Is AI 2027?
AI 2027 is the most detailed public forecast of the path to superintelligence ever published. It was released on April 3, 2025 by the AI Futures Project.
The team includes:
- Daniel Kokotajlo (lead author) — ex-OpenAI researcher who resigned over safety concerns
- Scott Alexander — author of Astral Codex Ten, one of the most influential rationalist writers
- Thomas Larsen — AI Futures Project
- Eli Lifland — #1 ranked RAND forecaster
- Romeo Dean — AI Futures Project
The scenario describes a concrete, month-by-month path from current AI capabilities to superintelligence by 2027, including infrastructure investment, agent deployment, AI-for-AI-research feedback loops, geopolitical dynamics, and security failures.
What You'll Find Here
- 48 tracked predictions — each with its own page showing the original claim, current evidence, counterevidence, and what would change our assessment
- Predicted timeframes — when the essay expects each prediction to materialize
- Status tracking — from "confirmed" to "not yet testable," with explicit reasoning for every judgment
- A transparent methodology — how we extract claims, how we score them, how we handle ambiguity
- Evidence you can inspect — every assessment is sourced
How This Works
The AI 2027 essay is a narrative. We break it into discrete, trackable claims and monitor them against reality. Some are quantitative (compute targets, benchmark scores), others qualitative (geopolitical dynamics, organizational decisions). We update when meaningful evidence arrives — not to chase news cycles, but to keep an honest record.
See our Methodology page for details on how we score predictions.
Independence
This is an independent project, not affiliated with the AI 2027 authors, any AI lab, or any forecasting organization.
Contact
Questions, corrections, or evidence we should look at? See the Impressum for contact details.
— Johannes Haus, Hamburg, Germany