@vatsalbaherwani

Thread

Scaling laws predict an LLM's pretraining loss, but not its capabilities. Abilities like in-context learning emerge abruptly and only past a certain scale. Our new paper traces this to one bottleneck: learning which tokens attention should focus on. 🧵<a target="_blank" href="https://arxiv.org/abs/26...

@andrewmccalip

Thread

Get paid to wait The Claude Code spinner might be the most watched line on Earth. So I turned it into an ad marketplace. Advertisers bid on it. You keep 50% of the money. Install the extension → get cash from ads. Introducing Kickbacks ...

@zodchiii

Article

Anthropic ran the numbers: unguided Claude Code succeeds just 33% of the time....

@AnatoliKopadze

Article

---...

@0xCodez

Article

You’re paying <b>$20–$200</b>/month for Claude and using it like a paid version of ChatGPT. <b>9 </b>out of 10 users have never set up automation that runs without them clicking buttons....

@jota_snchez

Thread

Una psiquiatra de Stanford avisa: la gente está más deprimida y ansiosa que hace 30 años porque vivimos enganchados al placer fácil. 8 consejos para resetear tu dopamina de forma natural: 1/ No tocar el móvil la primera hora del día. ...

@kingwilliam_

Article

two people open AI on a monday morning. same goal, same apps, same phone....

@gemchange_ltd

Article

> I had a swarm running. 80 agents on the same task, the kind where you can check the answer at the end. About a third of them were quietly garbage....

@Fefecilo

Thread

🧵 Esta época de mundial, es maravillosa para romper con esa falsa "hermandad letrinoamericana" promovida y exaltada por México, lo importante es reforzar la región de Sur América y lograr mayores afinidades entre las dos agendas regionales "el Pacto Andino y Mercosur" 🔽...