P

The man without qualities. History (record things for posterity!), wikis and sharing knowledge are important.

[P @Pier4r]

new pin it is

[P @Pier4r]

@LechMazur You mean creating an "hybrid" (fan-friendly) swiss? Yeah that will do I think, and could be more robust, then again UEFA got to that point pretty late

[P @Pier4r]

@LechMazur because otherwise with more games players get burned out or at risk of injury. A swiss format is terrible for the logistic of the viewers in the stadia.

I'd rather prefer leagues played over the years, but then again the team changes as well over the years

[P @Pier4r]

@LechMazur This is an old and not so easy to solve problem.

The constraint is: you want at most 7 to 8 games max for the winning team, and the entire thing running for around 1 month.

It is not easy to define another format (maybe the problem could become an LLM benchmark)

[Kashaf @noor36758]

the weekend really humbled him 💀

[nicolas, 30 ans @nic_carter]

I can’t believe Anthropic comparing their product to nuclear weapons 800 times backfired on them. I am shocked

[Sam Ashworth-Hayes @SAshworthHayes]

Updating the classics

[sdmat @sdmat123]

Anthropic

[ThePrimeagen @ThePrimeagen]

I heard the reason why Amazon reported on Fable to the government is because Fable was unable to fix how bad AWS's UI is.

[P @Pier4r]

https://t.co/XSJhHt4WPA

[Rhys @RhysSullivan]

last one

[Similarweb @Similarweb]

The explosion in app creation is not translating into a broader pool of successful apps.

Since February 2025, roughly 75% of newly launched Android apps failed to exceed 1,000 cumulative downloads, while only 2% surpassed 100,000 downloads.

The overwhelming majority struggle to gain meaningful traction with users.

[P @Pier4r]

They are optimizing for it, I don't buy that is "emergent" as other niche bench aren't seeing big jumps compared to competitors (but big jumps compared to opus)

[Epoch AI @EpochAIResearch]

Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.

[banteg @banteg]

updated fable eval scores

[An Engineer's Log @an_engineer_log]

@AnthropicAI

[P @Pier4r]

That seems to be quite an hard goal.

[Rob Lee @RALee85]

Ukrainian Defense Minister Mykhailo Fedorov said he aims to fill 30-50% of assault and infantry positions with foreigners.
https://t.co/rAdV1iVKZs

[Ed Zitron @edzitron]

Product so good they are literally giving it away

[Stephen Punwasi 🏚️📉🐈☃️ @StephenPunwasi]

lmao. Facebook’s down because…

*drumroll*

… they didn’t lint the program that sends its json, and now its malformed.

Another classic from the golden age of AI slop code.

[P @Pier4r]

Kudos for admitting and fixing the problem. SOTA is almost saturated.

But anyway, static benchmarks aren't going to cut it (though they are impressive milestones)
https://t.co/QpUX3IWqGI

[Epoch AI @EpochAIResearch]

FrontierMath: Tiers 1–4 (v2) is live.

We concluded an audit that addressed errors in 42% of problems. Rankings are similar but scores are higher across the board. The current leaders are GPT-5.5 (xhigh) with 85% on Tiers 1–3 and Google’s AI co-mathematician with 76% on Tier 4.

[math prof @mathematicsprof]

Have you ever wondered what major discoveries were made in mathematics exactly 100 years ago? This webpage answers. https://t.co/J0uoj8PWBr

[Ahmad @TheAhmadOsman]

To everyone defending Anthropic by mentioning how things are against their ToS: Cry me a river they trained their models on the entirety of humanity's knowledge