Podcast host Dwarkesh Patel published a long essay on November 13 subtitled "Why I now expect superintelligence by 2030." He writes that a few months ago he thought superintelligence was at least the better part of a decade away, and now thinks AIs as good as the best humans at AI research are only a couple of years, possibly a couple of months, away. His reasoning: AI has rapidly broken through verifiable domains like math, and AI research is itself largely verifiable. The post has about 820K views and 4,129 likes, quote posts from people like Dan Robinson spread the debate, and X lists it among trending stories. This piece covers one person's views and community discussion, not a verified forecast.

Key Takeaways

  • ✓Dwarkesh's post: "The avalanche that hit math is coming for AI development," linking the essay and a narrated video; about 4,129 likes, 346 reposts, 138 replies and roughly 820K views (post)
  • ✓The core claim: AIs as good as the best humans at AI research are "only a couple of years away, and possibly only a couple of months away"; automating AI research would compress so much progress into one year that AIs at least as good as top experts in every field could arrive by 2030, plausibly as early as next year
  • ✓Evidence he cites from public sources: METR concluded after three months inside Anthropic that its progress rate on well-scoped tasks is 2x a year earlier; an OpenAI September blog post said its heaviest 10% of researchers went from about $1,000 to over $7,000 a day of internal coding agents; Anthropic said Claude "leads" more than 26% of its AI R&D work as of August
  • ✓Dan Robinson quoted it saying the footnote on compute bottlenecks "hit me like a sledgehammer" (about 2,461 likes); Dwarkesh followed up that even without a full singularity, the intelligence explosion might compress 4 years of progress into one
  • ✓Skepticism is part of the thread: one commenter on the essay said this is "not what I am seeing at all" in daily work, and another asked whether agent spend is a good proxy for progress; these are opinions, not established facts
⚡

Key Decision Metrics at a Glance

TL;DR Verdict
🏆 Benchmark RankRank
No model benchmark in this story · see current rankings
Full Leaderboard→
💰 Cost & PlansCompare
Evidence he cites from public sources: METR concluded after three months inside Anthropic that its progress rate on well-scoped tasks is 2x a year earlier; an OpenAI September blog post said its heaviest 10% of researchers went from about $1,000 to over $7,000 a day of internal coding agents; Anthropic said Claude "leads" more than 26% of its AI R&D work as of August
Compare Plans↗
🛠️ Ecosystem & IDERules
X discussion · Dwarkesh Patel · Superintelligence
Setup IDE Rules↗
🧭

Turn your technical choice into a development budget

Compare 40 dev plans & simulate token costs vs $20/mo subscriptions

⚔️ Compare Model→💰 Compare plans & limits↗
🔗

Project Links & Resources

Direct Access
Direct access to official project resources and documentation
📝
Dwarkesh Patel: The avalanche that hit math is coming for AI development
www.dwarkesh.com/p/rsi
↗
🔬

In-Depth Technical Analysis

Background

Over the past few months AI has knocked down a string of long-open math problems, and whether AI can speed up AI research itself has become one of the hottest debates on X. On November 13, podcast host Dwarkesh Patel published the essay "The avalanche that hit math is coming for AI development," subtitled "Why I now expect superintelligence by 2030." The post has about 820K views and 4,129 likes, he added several follow-up threads on November 14, and X has grouped the debate into a trending story.

The argument

His main line: today's AIs are spiky and jagged, worse than humans in most domains but as good or better in ones with verifiable proxies like math, and AI research is largely a verifiable domain too. Once AI can automate AI research, it would compress a huge amount of progress into a single year. From this he concludes that AIs matching the best humans at AI research are a couple of years, possibly a couple of months, away, and that AIs at least as good as top experts in every field could arrive by 2030, plausibly as early as next year.

Numbers that matter

This is not a new benchmark. The figures in the essay come from public sources: METR found Anthropic's progress rate on well-scoped tasks is 2x a year earlier; OpenAI's heaviest 10% of researchers went from about $1,000 to over $7,000 a day of internal coding agents; Anthropic said Claude leads more than 26% of its AI R&D work as of August. The verifiable post metrics are about 4,129 likes, 346 reposts and 138 replies.

What developers should take from it

For developers, it works as a framing: the work that is easiest to verify automatically, with tests and eval sets, gets taken over by AI first. But this is Dwarkesh's personal forecast, and commenters on the essay say they see nothing like this pace in daily work. For real tool and model choices, rely on official releases and your own testing.

Action HubReady to adopt this in production?

Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.

⚔️ Launch VS Compare→💳 View Plans↗
Up NextX Debate: AI Commentator Chubby Says Claude Code and Opus 5.5 Have Been His Daily Driver for Two Months and GPT Is "Not Close," but Concedes Computer Use Is Still GPT's Edge; 150+ Replies Focus on Splitting Work by Task and the Same-Day Claude Login Outage
Read Now→