AIs can now often do massive easy-to-verify SWE tasks and I've updated towards shorter timelines

2026-04-06 · Source: AI Alignment Forum · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Expert, extended

Summary

An editorial analyst has significantly updated their AI timelines, nearly doubling the probability of full AI R&D automation by EOY 2028 to just under 30%. This revision is driven by unexpected performance from models like Opus 4.5, Opus 4.6, and Codex 5.2, which exceeded benchmarks and demonstrated the ability to complete massive, easy-to-verify software engineering (SWE) tasks that would take humans months to years. The analyst now expects AIs to achieve a 50%-reliability time horizon of years to decades on these "Easy-and-cheap-to-verify SWE tasks that don't require much ideation" (ESNI tasks) by EOY 2026. Key factors include observed superexponential progress on ESNI tasks, anticipated substantial training compute scale-up in 2026, and a greater-than-expected "scaffolding overhang" allowing significant improvements with straightforward scaffolding. This accelerated progress is also expected to speed up AI R&D itself.

Key takeaway

For research scientists evaluating AI capabilities and future development timelines, you should account for the observed superexponential progress in AI's ability to handle easy-to-verify, iterative software engineering tasks. This shift implies that AI R&D automation may arrive significantly sooner than previously projected, with a 30% probability of AI R&D parity by EOY 2028. Your strategic planning should consider the accelerating pace of AI progress, especially in areas where tasks can be highly decomposed and verified.

Key insights

AI progress on easy-to-verify tasks is accelerating superexponentially, shortening timelines for AI R&D automation.

Principles

• Iterative optimization drives AI performance on verifiable tasks.
• Scaffolding significantly enhances AI utility for complex tasks.
• AI R&D accelerates as AI becomes more useful for R&D.

Method

AIs can develop test suites and iteratively optimize solutions against them, enabling progress on well-specified, easy-to-verify tasks even with initial errors or poor judgment.

In practice

• Decompose large tasks into easy-to-verify components for AI.
• Implement robust scaffolding to guide AI through complex projects.
• Provide frequent, brief human feedback to correct AI prioritization.

Topics

• AI Timelines
• AI R&D Automation
• Easy-to-Verify SWE Tasks
• AI Model Performance
• Scaffolding & Orchestration

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• AIs can now often do massive easy-to-verify SWE tasks — AI progress on verifiable software tasks is accelerating superexponentially, shortening timelines for AI R&D automation.
• The next 36 months will be WILD — Expert consensus points to AGI/ASI by 2027-2028, driven by super-exponential AI progress and an irreversible industrial race.
• Leading AI expert delays timeline for its possible destruction of humanity — An AI expert has delayed his timeline for AI achieving autonomous coding and superintelligence, citing real-world complexities.
• Claude BROKE Wall Street Overnight... — AI progress is exponential, not cyclical, driven by consistent capability growth and strategic deployment models.
• Measuring AI Ability to Complete Long Software Tasks — AI's software task completion ability doubles every 7 months, potentially reaching month-long tasks by 2029.
• The loop of progress — Progress stems from relentless, iterative refinement through ideation, experimentation, and analysis.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Alignment Forum.