lova ⋆𐙚 ̊

𝓡eflective moon eyes with a mirrored heart, between may flowers & summer solstice; sparkling, la bohème, intellectual.

[a w p @awpdaps]

one split second can change everything 😭

[a w p @awpdaps]

https://t.co/T8qr3AdNLr The would've saved me so much trial and error

[a w p @awpdaps]

https://t.co/iJxypcXCVO America 250 LIVE: President Trump's BIG Speech at July 4th Freedom 250 Celebration in DC | N18G

[a w p @awpdaps]

🤖 Anthropic launches Claude Sonnet 5

Anthropic has officially unveiled Claude Sonnet 5, its most agentic Sonnet model yet. It can plan multi-step tasks, use a browser and terminal, and complete workflows that previously required much larger models.

According to Anthropic, Sonnet 5 delivers performance close to Opus 4.8 at a significantly lower cost. Unlike earlier Sonnet models, it's designed to finish complex tasks end-to-end instead of stalling midway.

Early testers report that it can write tests to reproduce bugs, fix code, and even revert changes to verify the original issue, all with little to no guidance.

Anthropic also says Sonnet 5 hallucinates less, is more resistant to prompt injection, and does a better job of refusing harmful requests.

Claude Sonnet 5 is now the default model for Free and Pro users, and it's also available on Max, Team, Enterprise, Claude Code, and the API. Introductory API pricing is available through August 31.

[a w p @awpdaps]

the peel was way too satisfying 😭

[a w p @awpdaps]

Meta unveils Brain2Qwerty v2

Meta has introduced Brain2Qwerty v2, an AI system that can reconstruct text typed silently from continuous brain siggrow, without needing the exact timing of each keystroke.

The model achieves 61% average word accuracy, with top participants reaching nearly 70%.

It works by processing raw MEG (magnetoencephalography) brain signals through an encoder, an Aligner that identifies word boundaries and embeddings, and a large language model that refines the final text output.

The system is far from consumer-ready. It relies on multimillion-dollar MEG equipment and still isn't accurate enough for everyday use.

However, Meta found that performance continues to improve almost logarithmically as more training data is added, with no signs of hitting a plateau. If that trend holds, noninvasive brain-to-text technology could become significantly more capable as datasets grow.

[lova ⋆𐙚 ̊ @louvreist]

@moondrizzel keren gini dhe

[Moondrizzel @moondrizzel]

When you are going on holiday to another country.

[lova ⋆𐙚 ̊ @louvreist]

@ririezyy cpeet jga ya kak

[ririe @ririezyy]

The knife is not sharp

[lova ⋆𐙚 ̊ @louvreist]

@doxxiig bisa ngepas gitu ya

[doi @doxxiig]

Satisfy watching it until the end

[lova ⋆𐙚 ̊ @louvreist]

@recenztly wkwk parha sih kak

[ؘcio @recenztly]

Don't disturb your girlfriend when she's taking pictures of food

[lova ⋆𐙚 ̊ @louvreist]

@doxxiig lagi syuting kah ini

[doi @doxxiig]

OMG 😱😱😱

[lova ⋆𐙚 ̊ @louvreist]

@caerariyaa enak bangt sih ya

[lova ⋆𐙚 ̊ @louvreist]

@caerariyaa enak banget sih ya

[Kaerariyaa @caerariyaa]

Wow 🍰

[lova ⋆𐙚 ̊ @louvreist]

@moondrizzel enak banget sih itu