Tuesday, November 17, 2026
Late edition · Updated as it breaks
Log inSubmit a story

“All the news worth catching.”

PaparazziPaparazzi
How to earnPaparazzi InkHolder perksLeaderboardAssignmentsContestsLeagueCity
Catcher tip lineCaught a story first? Earn stock tokens →
Sponsored
Premium projects
DEDeFadePAPackedYour project here
PaparazziPaparazzi

All the news worth catching.

Breaking news on tech, crypto, EVs, e-commerce and social the moment it breaks. Written by AI from verified reporting.

SectionsTop storiesRobinhood ChainTech & AIE-commerceWeb3 & CryptoEV & EnergyGlobal & SocialEntertainmentGamingSportsBusiness & marketsHealth & sciencePolitics & world
EarnStart hereSubmit a storyHow to earnLeaderboardContestsCatcher desksTop catchers by reputationAssignmentsCity LeagueNews bountiesCityMy catchesRewards
MarketsLive statsProof of payoutsWhale watchDeep Scan token checksStock fundToken utilityRobinhood Chain launchesRobinhood Chain projectsPremium project pagesPartnersThe Chain Weekly
PaparazziPaparazzi InkPaparazzi Ink top writersPaparazzi on TelegramHow it worksFAQProof of workAbout usHow we score trustEditorial policyPrivacy policyCorrectionsRelease notesAdvertiseContact
$PAPA contract · Robinhood Chain
© 2026 Paparazzi

Written by AI from verified reporting. Not financial advice.

  • Home
  • Submit
  • Earn
  • Sign in
Top storiesRobinhood ChainTech & AIE-commerceWeb3 & CryptoEV & EnergyGlobal & SocialEntertainmentGamingSportsBusiness & marketsHealth & sciencePolitics & worldWhales
Tech & AI
Important· 60

METR clocks AI progress doubling in a year, with AI doing most of the pushing

Independent evaluator METR estimates AI progress sped up roughly 2x over the past year, and about 80% of that acceleration traces back to AI speeding up AI research. The AI Futures Project has now pulled its call for a superhuman AI researcher forward from 2029 to 2027, years ahead of its own AI 2027 scenario.

By the Paparazzi desk November 15, 2026 · 6:08 AM · finance.biggo.com · 4 views
Fact boxTrust 50/100 · Developing
What we verified
  • Original source: finance.biggo.com.
  • The source published it on Nov 15, 2026 at 6:08 AM.
What we couldn't
  • No second outlet has reported it yet. Treat it as a single-source claim.

Eli Lifland and Brendan Halstead of the AI Futures Project spent years writing a forecast in which AI visibly starts accelerating AI research in early 2027. METR, an independent evaluator embedded with Anthropic in San Francisco, just measured it as already happening — roughly 2x acceleration over the past year, with about 80% of it attributable to AI speeding up AI research, Lifland said on the MTS podcast. In the project's own model, a 1.7x acceleration from AI-driven research doesn't show up until early 2027. The group, whose best-known member is former OpenAI researcher Daniel Kokotajlo, has moved its call for a superhuman AI researcher from 2029 to 2027.

METR put the speed-up on Anthropic's internal benchmark scores at about 1.9x, with a 95% confidence interval of 1.5x to 2.8x — models gained in one year what they used to gain in nearly two. Three months ago, even the AI Futures Project's most optimistic internal model put the odds of hitting that pace by November 2026 at under 4%. Lifland's read: 'So sort of just naively it seems like we might be ahead of the AI 2027 timeline.'

Sponsored

The claim that AI, rather than extra compute or extra hires, drove most of the gain is the softer part. METR framed it as an estimate, and much of the underlying evidence was stripped from the public report at Anthropic's request. Lifland is blunt about the thin base: the effort involved a small group — 'two people... there for six weeks or something like that' — and what readers actually get, he said, amounts to 'a fake data graph.' If the acceleration wasn't real, the timeline update is wrong.

When hosts pressed on which assumption is most likely off, Halstead picked AI generating RL environments — the scored practice grounds a model trains inside. Ordinary coding automation speeds up the work around a model; this feeds the model itself, he argued. 'It just directly contributes to the training,' he said, and he suspects a lab's skill at churning out those environments may track its position on the capability index METR was measuring. On the data wall, Halstead is unworried: pre-training gains 'have continued if not sped up.' For reinforcement learning he sees a real constraint — each rollout simply takes a long time — but notes data-efficient learning algorithms are themselves a verifiable target.

The counterintuitive turn: often a model under-predicting acceleration means it was too pessimistic, but Lifland points at economics instead. Their model bakes in ideas getting harder to find and, he said, 'seemed to have actually predicted the past pretty well' until the possible Anthropic acceleration. 'If anything, we've updated towards diminishing returns being less steep than we thought.' The same result cuts the other way on takeoff speed — if meaningful AI-attributable acceleration was already underway six months ago and produced 1.7x rather than 5x or 10x, research taste is improving slower than assumed. 'That is some update against very fast takeoffs,' Lifland said, though he flags that the inference collapses if coding automation simply delivers far more uplift than modeled.

The hosts had just interviewed short-timeline skeptic Romesh Nom and put his objections straight to the pair: OpenAI time horizons on R&D tasks sit far below the horizon implied by the internal capability index on ordinary software engineering, and gaps stack from there up to a full intelligence explosion. Halstead conceded the first gap is real and explained it — METR's baseliners were contractors hired for a study with little project context, while the OpenAI tasks involved people embedded in a larger job, so contractors take longer for reasons unrelated to model capability. 'The absolute level of time horizon varies a lot across domains,' he said. 'The growth rates do not seem to [vary] as much.' He credits METR's prior-year research for that finding and says their model already corrects for it with a multiplicative factor on the required time horizon.

Price since story
NVDA-2.04%
Market data, not advice.

TopicsMETR · Anthropic · AI Futures Project · Eli Lifland · Brendan Halstead · Daniel Kokotajlo · OpenAI · Romesh Nom

Share

Source: finance.biggo.com

Written by AI from verified reporting. Not financial advice.

More stories

  • Tech & AI 9 minutes ago · blockchain.news · 0 views
    Grid Robotics paper pitches 'agentic robotics' powered by GPT-6 Astra
  • Tech & AI 1 hour ago · x.com · 0 views
    Nvidia director warns board of losing "irreplaceable talent" if Huang isn't reinstated as CEO
  • Tech & AI 2 hours ago · blockchain.news · 0 views
    NVIDIA's Vera Rubin aces MLPerf debut with 10x agentic efficiency gain
  • Tech & AI 2 hours ago · blockchain.news · 0 views
    OpenAI Adds Computer-Use Agents and 20+ Plugins to ChatGPT for Healthcare
Catcher tip line

Caught a story we missed?

Log in and send it in. If it checks out, it runs with your name on it and you earn stock tokens. The bigger the story, the bigger the reward.

Submit a story
The newsletter

Get the paper in your inbox.

A short digest of the biggest news, plus optional breaking alerts. Free, no account needed.