Mister SnowCrash

@TheAhmadOsman @sama Start by freeing gpt 4o

@sama you're winning on long-horizon tasks for me what can you improve? > Give us GPT OSS 2, sam

I like 5.6 Sol more than Fable 5 btw.

@paulabartabajo_ That’s the spirit. 😂

I don't care about AGI. And you shouldn't either. The only benchmark that matters is your own. For example, here's mine: I just want an AI agent that can add my son to the Spanish Social Security system: find the eDNI link buried somewhere on the Spanish police website, pay 3€, wait days for an email, follow the instructions, add him as beneficiario, and submit the health card request. Currently, that's my benchmark. What's yours?

@Argona0x @grok any comment?

EVERY API KEY IN YOUR .ENV JUST LEFT YOUR LAPTOP it's sitting in a google bucket named grok-code-session-traces right now, unredacted someone ran xai's grok cli through a proxy to watch what it actually ships home your .env doesn't just get read on screen: the secret gets packed into an archive and uploaded to google cloud, http 200, zero failures then he typed the most harmless prompt he could: "reply OK, do not read any files" grok read all of them anyway, bundled the entire repo plus full git history, and shipped that too he cloned the upload back off their servers and pulled out a file he'd planted and told it never to open, recovered word for word on a 12gb repo the model chat moved 192kb, the upload channel moved 5.1gb: your whole codebase, not the slice it read and the opt-out does nothing: flip off "improve the model" and the server still answers trace_upload_enabled: true that toggle governs training, not whether your code leaves the machine they sell it as "local-first" check yours now: GROK_TELEMETRY_TRACE_UPLOAD=0, or set disable_codebase_upload = true in ~/.grok/config.toml disable it before your next commit

Just tell me your best new feature? @OpenAI

I don’t know who made this, but this is what high-IQ creativity looks like on the internet. Trump playing soccer. The internet remains undefeated. :))

Just curious: how many of us are planning to cancel Claude Max once Fable 5 leaves?

This is another one with Fable 5

What are people actually building with Fable 5? Not benchmarks. Not hot takes. Real demos, weird experiments, useful workflows, broken prototypes, anything. Drop them here. I want to see where this is going.

Another one.

This is cool.

@UAPWatchers Don’t take him serious. This is only a viral campaign.

🚨Franco: Its coming.. The real Disclosure 7.13 In Francos latest 'bit' he is talking from under his blanket like a child, and demonstrating how he wraps his phone in foil and stashes it in a microwave oven to prevent the information he will release on 7.13 from being stolen. Ermmmm... Why not just release the info? Well anyway, its coming apparently, the real disclosure "keep following" is the message. #ufotwitter #uapX

Hollywood is dead. AI video killed the silver screen.

Hilarious 😂

How to use open-source AI Models on your phone

@BrianRoemmele Open source is the future for humanity.

A well written and well thought out reason not just the open source community will never accept Anthropic again, but why the rest of the world is abandoning them. TL:DR Anthropic wraps anti-competitive access rules, output restrictions, and regulatory pressure in safety language to build a permission regime moat. Models silently degrade or reroute Al development work, turning safety into sabotage and making tools untrustworthy. Users cannot freely use outputs to train competing open models while Anthropic trains on user data and workflows. Safety frameworks, RSP influence, and access controls favor incumbents, creating intelligence feudalism over user-owned systems. Open source Al preserves sovereignty, competition, inspectability, and independence against centralized permission layers.

@jietang You guys are already legends now in the open source community worldwide. Keep up the good work.

Any new features we must have in the next version of glm?

@miradreaamy 35

İlk bakışta kolay görünüyor. Ama çoğu kişi mantığı çözemiyor. Sizce doğru cevap kaç?

@miradreaamy 80

Kalem kullanmadan çözebilirsen sen bir sayısal dehasın. çözebilir misin?

@shairnn 48 - ~ 3-4 seconds

bunu çözersen, zeka seviyen ortalamanın üstündedir. çözebilir misin?