

2026 in Review, and Predictions for 2027
A year ago almost none of this was obvious. In September an AI system outscored the best human contestant at the International Olympiad in Informatics. Open-weight models closed the gap to the frontier to a matter of months, and Meta walked away from open weights entirely. Anthropic overtook OpenAI on quarterly revenue and took the lead in enterprise. Agents went from demo to deployed. World models went from a niche research term to the largest seed round in European history.
A good share of what everyone was certain about in January turned out to be wrong.
We go through what actually changed in 2026 and what it means for the companies being built now. Then we make seven predictions for 2027, specific enough to be wrong, and we will grade them in front of you next December.
20 minute presentation, 30 minutes for questions.