gramanoid

dangerously skipping permissions.

[gramanoid @alexgrama]

@Dexerto Why is he doing side quests?

[Grace Kay @graceihle]

Scoop: Cursor is working on a competitor to Claude Cowork.

The new Cursor product, internally referred to as Sand, is its first product aimed at more casual AI users rather than developers.

[gramanoid @alexgrama]

@testingcatalog Insane feature

[gramanoid @alexgrama]

@BrainsAndTennis none of these words are in the Bible

[gramanoid @alexgrama]

@wgussml omg is that tibothy

[gramanoid @alexgrama]

@DavidOndrej1 he's baiting for engagement

[gramanoid @alexgrama]

@jun_song they got mogged by 5.6 and Fable and now they're back on the RL bench

[gramanoid @alexgrama]

@samsheffer @GoogleAIStudio since when is this a benchmark? why is the bar so low? how is this a "🤯" moment?

[gramanoid @alexgrama]

@harshilmathur @alexandr_wang you got WANGd bro

[Abby Grills @AGrillz]

claude just called me a cunt

[gramanoid @alexgrama]

@thsottiaux can we store it instead ? you already reset yesterday. love u

[gramanoid @alexgrama]

@OpenAIDevs @OpenAI @thsottiaux @ajambrosino

hi

1. any reason we cannot choose our preferred model in the Powerpoint / Excel extensions even though we can do it in the Chrome extension?
2. any reason the context size in codex is 353K while via API it's 1 mil? why can't we make use of the full context size?
3. @sama mentioned at a certain point that Sol will be available at 750t/s via @cerebras . the fast mode now still only has the 1.5 modifier

[gramanoid @alexgrama]

@RohdeAli it was one of the first and main announcements. how is it easy to miss?

[gramanoid @alexgrama]

@ryanbrewer say sike rn

[OpenAI @OpenAI]

Meet the Wishingrads.

A family running a cereal business from their dining room with GPT-5.6.

[gramanoid @alexgrama]

@claudeai not now lil bro, gpt 5.6 is launching

[gramanoid @alexgrama]

@jacobandreou loosest and longest assignments sounds like the shits I take after Chipotle

[gramanoid @alexgrama]

@thekitze what the fuck is the problem now?

[anton @abacaj]

the three paths: become a wrapper, train big models, or sell compute

[enzo @enzo_gte]

It's kind of hilarious that the new meta for AI is:

1. US frontier lab trains closed-source model.
2. Chinese company distills said closed-source model and open-sources it.
3. American company RLs open-source model to get to closed-source model performance at a fraction of the price.

Capitalism at its finest.