AI R&D progress multiplier reaches 1.5×

What AI 2027 Predicted

The scenario predicts that AI contributes meaningfully to its own research and development, reaching a 1.5× multiplier — meaning AI research proceeds 50% faster than it would with human researchers alone. This is a precursor to the recursive improvement dynamics described later in the timeline.

How We Track This

We monitor:

• PostTrainBench and other AI R&D automation benchmarks
• Lab reports on AI contribution to research productivity
• METR pilot studies on AI research acceleration
• Academic papers measuring AI’s impact on ML research

Current Evidence

AI Futures’ self-grading (Feb 2026) assessed AI software R&D uplift as “behind pace” relative to the scenario’s predictions, though they note the gap may be closing. PostTrainBench (March 2026, arXiv:2603.08640) provides the first standardized measurement: GPT-5.1 Codex Max post-trained a Gemma-3 model to 33% on GPQA vs 31% for a human-trained version — a modest improvement that falls short of demonstrating a 1.5× multiplier. METR’s pilot study found ~25% of forecasters think >3× AI progress acceleration is likely if AI achieves researcher parity by 2027. Karpathy’s AutoResearch agent (630 lines) runs autonomous ML experiments, though as a proof-of-concept rather than evidence of systematic R&D acceleration. All frontier labs now have internal AI research automation infrastructure, though the productivity impact remains difficult to quantify.

Sources:

• Grading AI 2027’s 2025 Predictions — AI Futures Project
• PostTrainBench: Can LLM Agents Automate LLM Post-Training? (arXiv)
• Measuring AI R&D Automation (arXiv:2603.03992)
• Forecasting Impacts of AI R&D Acceleration — METR
• Karpathy Built an Autonomous AI Research Agent

Counterevidence & Limitations

• The PostTrainBench result (33% vs 31%) is a marginal improvement, not a clear 1.5× multiplier — it’s unclear whether this represents meaningful R&D acceleration or noise
• The AI Futures authors themselves assess R&D uplift as “behind pace,” which creates tension with an “on-track” status here
• Measuring the actual R&D multiplier is inherently difficult — most evidence is qualitative, and labs have incentives to overstate AI’s contribution to their research
• The “recursive improvement” may be more gradual than exponential — current evidence is more consistent with incremental tooling improvement than a step-change in research productivity
• Much of the current AI R&D automation covers implementation tasks (coding, testing), not the harder parts of research (hypothesis generation, experiment design, research taste)
• METR’s RCT found experienced developers were 19% slower with early-2025 AI tools — while tools have improved since, this result should temper claims of large R&D multipliers

What Would Change Our Assessment

• Upgrade to “confirmed”: PostTrainBench or similar showing AI matching human researchers on complex tasks
• Downgrade to “behind”: If the 1.5× threshold seems unlikely before late 2026

Update History

Date	Update
2026-04-13	No confirmed evidence of 1.5× R&D multiplier at frontier labs. METR simpler timelines model (Feb 2026) discusses measuring “uplift fraction” and acknowledges difficulty of measuring real-world R&D productivity gains (METR). Academic literature (GovAI, March 2026, arXiv:2603.03992) notes “it is unclear how directly such results translate to productivity boosts given real-world integration frictions.” The most advanced model had 80% success rate on tasks taking human expert coders 1h10m, but benchmark-to-productivity translation remains unproven. No lab has publicly claimed 1.5× R&D speedup from AI tools. Prediction remains behind.
2026-04-02	AI Futures Project Q1 update: AI company researchers “doubling down” on near-term automated AI R&D timelines in private discussions. Kokotajlo: “Rather than walking back their predictions, they are doubling down, both in public and in private.” The authors consider this some evidence (though far from conclusive) that the multiplier may be closing faster than external measurements suggest. Source: LessWrong
2026-03	AI R&D automation infrastructure in place at all major labs. Current estimates suggest ~1.1-1.3× multiplier, approaching but not yet reaching the 1.5× target.
2026-02	METR experiment redesign update (Feb 24): “developers are more sped up from AI tools now — in early 2026 — compared to our estimates from early 2025.” Directionally positive but not yet quantified.
2026-01	AI Futures clarification post (Jan 27): “AI software R&D uplift: behind pace.” METR RCT and gap between benchmarks and real-world productivity remain primary evidence. Anthropic 2x internal coding uplift (self-reported, non-randomized) is best evidence in favor.
2025-12	AI Futures Project Dec 2025 model update explicitly cited R&D multiplier as key reason timelines shifted longer: “less pre-SC AI R&D automation effect, due to fixing a bug and revising the interpolation method.” AI labs report meaningful but modest acceleration of internal R&D workflows through AI tools. Estimated multiplier approaching but not yet at 1.5×.
2025-08	METR’s August 12 research update confirms and elaborates its July finding: a controlled study found experienced developers working with early-2025 AI tools were 19% slower, not faster. METR attributes this to systematic overestimation of AI utility on tasks requiring human judgment. This is direct contrary evidence to the 1.5x multiplier prediction. The prediction’s mechanism (benchmark performance translating to R&D productivity) is not confirmed.
2025-07	METR publishes RCT (July 10): experienced open-source developers using Cursor Pro with Claude 3.5/3.7 Sonnet took 19% longer to complete tasks than control group. 16 developers, 246 real GitHub issues. Developers perceived 20% speedup while experiencing slowdown. Significant counterevidence against near-term 1.5x R&D multiplier.