• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Anthropic shares three measures to track AI development

    Anthropic says its internal snapshot covers how much AI research is done by AI, how well AI agents are overseen, and how computing power is allocated.

    2 Sources, 20d ago, first seen 20d ago

    TLDR

    Anthropic shared three measures covering AI’s role in research and development, agent oversight, and computing resources, along with a snapshot from inside the company. It says any frontier developer could publish the same measures and third parties could verify them. Anthropic’s stated goal is to narrow the information gap between labs and the public, giving society more insight as it considers the pace of advanced AI development.

    Combined views

    —

    2 Sources, first seen 20d ago

    Combined views

    —

    2 Sources, first seen 20d ago

    — likes
    — likes
    — comments
    — saves
    — reposts
    — comments
    — saves
    — reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    2 Sources

    zhod@zhodonxNEW UPDATE: Anthropic wants to measure how much of the next generation of AI is actually being built by AI itself. And more importantly, whether humans can still keep up with the work. I’ve talked a lot about how ridiculously fast model releases are becoming. Grok 4.5 launched in July. Just 27 days later, we got Grok 4.6. Anthropic went from Opus 4.7 to 4.8 in just six weeks, then released Opus 5 less than two months later. And behind all these releases, AI is already helping researchers build better AI. Claude is already writing code, running experiments and taking on increasingly complex research tasks inside Anthropic. Plus, their engineers are now shipping roughly 8x as much code per quarter as they did in previous years. Anthropic just published a new report trying to make the progress behind all of this more visible. Basically, they’re looking at three things: ⟣ How much AI research and development is actually being done by AI itself? ⟣ How well can humans oversee the agents doing that work? ⟣ And how much computing power is going into it? I believe this insanely interesting because as models get better at research, they actually can help build even better models. But eventually, we could reach a point where AI is producing new research faster than humans can meaningfully review it. So it’s typically a human problem. But anthropic assures that we haven’t reached that stage yet, and humans still play an important role in deciding which research directions are worth pursuing. But I think the bigger issue here is visibility. Right now, most of us are judging AI progress through model releases, benchmarks and whatever the labs choose to tell us. Meanwhile, a growing amount of the actual work behind those releases is already being done with AI. And if we ever reach a point where AI is significantly accelerating its own development, I’d want to know: ➥ How quickly that’s happening. ➥ How much of the work humans still understand. ➥ Whether the people overseeing it can actually keep up. That’s what Anthropic is trying to make measurable here. And tbh, with all the recent conversations about slowing frontier AI down, I think every major lab should eventually be expected to report this stuff. Just so we don’t recreate a second ultron.20d
    Temple Peter@TemplePeter2026LATEST 🔥: Anthropic says Claude now leads 26% of its own AI research and development, up from nearly zero at the start of the year.20d

    2 Sources

    zhod@zhodonxNEW UPDATE: Anthropic wants to measure how much of the next generation of AI is actually being built by AI itself. And more importantly, whether humans can still keep up with the work. I’ve talked a lot about how ridiculously fast model releases are becoming. Grok 4.5 launched in July. Just 27 days later, we got Grok 4.6. Anthropic went from Opus 4.7 to 4.8 in just six weeks, then released Opus 5 less than two months later. And behind all these releases, AI is already helping researchers build better AI. Claude is already writing code, running experiments and taking on increasingly complex research tasks inside Anthropic. Plus, their engineers are now shipping roughly 8x as much code per quarter as they did in previous years. Anthropic just published a new report trying to make the progress behind all of this more visible. Basically, they’re looking at three things: ⟣ How much AI research and development is actually being done by AI itself? ⟣ How well can humans oversee the agents doing that work? ⟣ And how much computing power is going into it? I believe this insanely interesting because as models get better at research, they actually can help build even better models. But eventually, we could reach a point where AI is producing new research faster than humans can meaningfully review it. So it’s typically a human problem. But anthropic assures that we haven’t reached that stage yet, and humans still play an important role in deciding which research directions are worth pursuing. But I think the bigger issue here is visibility. Right now, most of us are judging AI progress through model releases, benchmarks and whatever the labs choose to tell us. Meanwhile, a growing amount of the actual work behind those releases is already being done with AI. And if we ever reach a point where AI is significantly accelerating its own development, I’d want to know: ➥ How quickly that’s happening. ➥ How much of the work humans still understand. ➥ Whether the people overseeing it can actually keep up. That’s what Anthropic is trying to make measurable here. And tbh, with all the recent conversations about slowing frontier AI down, I think every major lab should eventually be expected to report this stuff. Just so we don’t recreate a second ultron.20d
    Temple Peter@TemplePeter2026LATEST 🔥: Anthropic says Claude now leads 26% of its own AI research and development, up from nearly zero at the start of the year.20d