LessWrong (30+ Karma) [Linkpost] “Q1 2026 Timelines Update” by Daniel Kokotajlo, elifland, bhalstead
Apr 3, 2026
Quarterly updates on AI timelines and takeoff speeds, driven by new model data and a faster METR time horizon. Discussion of revised forecasts that push key AI milestones earlier, including changes to automated coding timing. Examination of market signals and rapid agentic coding adoption as evidence leading to updated median forecasts.
AI Snips
Chapters
Transcript
Episode notes
Quarterly Timeline Updates Shifted Forecasts Earlier
- The AI Futures team is updating timelines quarterly and moved forecasts notably earlier after rapid recent progress.
- Daniel's Automated Coder (AC) median shifted from late 2029 to mid-2028 driven by METR v1.1 and new model evaluations like Gemini 3 and GPT‑5.2.
METR Doubling Time Revised To About Four Months
- The team revised METR doubling time estimates much faster, from a 5.5‑month median to ~4–4.5 months.
- This revision relied on METR v1.1 showing faster trends, new model time horizons continuing a fast 2024+ pace, and reanalysis of existing data.
OPUS 0.6 Prompted Much Faster AC Requirement Estimate
- Daniel cut his estimate for the 80% time horizon requirement for AC from three years to one year after OPUS 0.6's performance.
- Progress in agentic coding over the last 3–5 months exceeded expectations, especially due to OPUS 0.6's impressiveness.
