AI Futures trims AGI timelines, adds coding uplift and revenue anchors

AI Futures, the group behind the AI Futures Model at aifuturesmodel.com, published its Q2 2026 Timelines Update, introducing major methodological changes. Previously, forecasts of when an "Automated Coder" (AC) — an AI so capable that top AI firms would rather fire human engineers than forgo it — were anchored to METR's coding time horizon measurements. That approach had two critical flaws: ambiguity about what time horizon actually corresponds to AC, and extreme sensitivity to assumptions about superexponential growth in time horizons.

To address this, the team added two new anchoring methods: coding uplift (the observed speedup AI provides to software engineers) and AI company revenue. Coding uplift is now their preferred approach, since the AC threshold is far less uncertain and superexponential assumptions matter much less. Surprisingly, all three methods — time horizon, uplift, and revenue — yield very similar AC arrival dates, which the authors interpret as a sign of forecast robustness. The updated interactive model lets users explore each anchoring method via dropdown menus. The team also re-evaluated AI 2027's predictions, finding real-world progress is tracking at 70–90% of that forecast's pace. Consequently, overall timelines shifted slightly shorter, and the model now accounts for the training time required to apply software improvements, lowering the probability of extremely fast takeoffs. Individual authors assigned weights to each method, and the aggregate forecast is available at the provided link.