-‘

@LinusEkenstam You love paid promotions, dont you? ;)

You spent months engineering around context collapse. Better prompts. Smarter chunking. Careful retrieval. I've seen this before. Teams doing a lot of work, just so that one update later, all of that work gets obsolete. This is one such update. Sub-quadratic sparse-attention architecture, thats a mouthful, but 12 million token context window!?

@hasantoxr Love paid partnership…

This is the biggest LLM breakthrough in years. SubQ IS the first frontier model with a fully sub-quadratic sparse-attention architecture (SSA) + a 12 MILLION token context window. The numbers are insane: → 52x faster than FlashAttention at 1M tokens → Less than 5% the cost of Opus → Nearly 1,000x less compute by ignoring the 99% of token relationships that don’t matter No more quadratic waste. No more context hacks. Early access is open right now (plus their new coding agent SubQ Code). This is how LLMs actually scale next. Get it → subq.ai

@OunkaOnX @MiloradDodik this you?

Trump’s ‘gold card’ visas were supposed to solve the $39 trillion national debt. They’ve sold only one

@Tazerface16 Most dont, even worse some think they are sentient 🤣

People understand that LLMs aren't actually "thinking," right?

@andrewchen Haha WHAT to build is decided by someone who has money and not by somenone who flips states in jira tickets

bullish on the PM role quietly becoming the most important role in tech again when anyone can build, the person who decides WHAT to build becomes the bottleneck

@DeepakNesss I would be more afraid of mossad which already has all your data than Chinese

Genuine question: If I use the DeepSeek V4 Pro API via OpenCode, what data could the Chinese mine from me at worst? My entire codebase? Basic info about my computer? Just trying to understand how exposed I am in terms of privacy and security.

@RobertJBye Y but even small model can do this! Whats the advantage then of using opus giga max

This use case demo is 🔥 CAD has so many repetitive processes, so using Claude for these will be incredible!

@satyanadella I'm 101% sure that 99.999% licences wont be used xD

Super to see Accenture roll out 740,000+ M365 Copilot seats—our largest deployment to date! news.microsoft.com/source/feature…

36,000 engineering students from IT programs were tested in India. 95% were determined to be unfit for actually working in the field. Only 1.4% could write functionally correct and efficient code. India has a massive issue with cheating and scams in education. Fake degrees

@DearS_o_n 2 weeks until Iran has nuclear weapons

Name a huge scam that has been normalised?

@BFitzGab @TheMaineWonk @KimDotcom He deleted it… you trust llms too much

@TheMaineWonk @KimDotcom It’s bullshit …

What the fuck is this?

@yasinaktimur Nice, opened for 10 seconds and closed.

🚨 son dakika : GODSPECTIVE uygulaması tüm dünya’ya sızdırıldı. Artık tüm dünyada neler olup bittiğini canlı canlı yukarıdan izleyebiliyorsun. github.com/koala73/worldm… quanqiuchongtu.com

@heart_ Sometimes they ask AI whats their favorite planet 🤣

I still have no idea what people do all day with their $200/mo “AIs” except making dashboards and cleaning e-mails

@claudeai To be honest it looks like its generating themes from themeforest.net

Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude. Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.

@petergyang We tried locally 35B for coding. To be honest, its unfortunately useless

What is the sweet spot in open source model size? Are 35B models enough for local agentic workflows? Trying to decide how much RAM I need in a new computer.

@ElxMapping He should write a book about it 🤣🤣

visits the pope → pope dies leads Iran negotiations → talks collapse flies to Hungary to prop up Orbán → Orbán loses in a landslide Man’s got a streak.

@gfodor @kenwheeler Tell that to the guys who wrote attention is all you need paper…

@kenwheeler Nobody knows how AI works, anyone who says they do is an idiot

i deeply and i mean deeply dislike this man. with every fiber of my heart. and i don’t believe this for a second.

@TeksEdge That’s just benchmark tuning. I asked gemma to give me craftjs component, got plain react 🤣

❓ Interesting Benchmark: While Gemini 3.1 Pro sits at #9 on BridgeBench's UI ranking, its own Free and Open Source Gemma 3 31B MoE ranks #4.

@greg16676935420 Land is not fucking mspaint dude

This might be a stupid question but why don’t they just dig a new route?

@amjadt25 hahahhahahha who asked you?!

As long as the UAE is not in Islamabad, JD Vance and the Islamic regime in Iran will fail in Pakistan. The UAE does not, and will not, accept the Islamic regime in Iran controlling access to the Strait of Hormuz, nor possessing nuclear weapons, ballistic missiles, drones, or any capabilities that threaten regional security. This regime must issue a formal apology to the UAE for its aggression, acknowledge its actions, provide full compensation for any harm caused to individuals and institutions, and guarantee non-repetition. Otherwise, what comes next will be devastating for this terrorist regime.