Byakatonda Marvin

@SkyNews Is that some sort of message to the new Prime minister

BREAKING: Russian warship carries out live-fire weapons exercise 46 miles off UK coast, MoD says news.sky.com/story/russian-…

@OpenAIDevs Bring some of these events to Africa

Codex ambassadors are taking Build Week around the world 🌍

@chamath frames the AI price war as intelligence becoming a commodity like oil. When rival labs can sell the same “barrel” of tokens for radically different prices, expensive infrastructure bets face a brutal question of who captures the downstream profit?

Germany has advanced legislation that would make publicly denying Israel's right to exist or calling for its elimination a criminal offense punishable by up to five years in prison. The bill has been approved by the Bundesrat (Germany's upper house) and now moves to the Bundestag for consideration. If passed, it would expand Section 130 of Germany's Criminal Code beyond Holocaust denial to also cover public denial of Israel's right to exist. The proposal would apply to public statements, demonstrations and online speech, but only where the conduct is capable of promoting antisemitic violence or hatred. German officials said existing laws on incitement and support for terrorist organizations are insufficient and pointed to a sharp rise in antisemitic incidents, from 1,957 in 2020 to 8,627 in 2024. The proposal does not ban criticism of Israeli government policies or legitimate academic and artistic debate. The proposed bill is strongly supported by Christian groups in Germany, which argue that the Israeli state is crucial for the protection of Christian Arabs in the Holy Land.

Dario Amodei says the viral “AI will wipe out half of entry-level white-collar jobs” clip was never meant as a three-second doom soundbite. His full point is that disruption could be enormous, but policy responses from enterprise adaptation to taxation and macroeconomic action must be designed now.

@polynoamial I think this is the response to the behavior of GPT-5.6 sol deleting / wiping computers for some users. I hope this is permanently fixed by OpenAI as its a big risk for users who dont back up their work

Long-running models can solve hard open-ended problems, but their persistence can create safety risks that shorter-horizon evaluations miss. We’re sharing what we learned from studying a long-running model, and how those findings are shaping our approach to evaluations, alignment, monitoring, and user control. openai.com/index/safety-a…

@elonmusk Beautiful to see grok doing useful work in excel, I hope the 2T models will be all round better. I actually dont know why SpaceX stock has been in a continuous market drop the past 4 weeks, now it has lost nearly 50% of its peak value

Grok for Excel is now live

@sama Hey Sam, is there any chance that OpenAI will release an Open source model this year, its been a while since your last release?

it is good now!

@stkms0011223344 @chamath Your way of seeing things

@Marvinma12 @chamath Because they're grifters mooching off taxpayer dollars 🤑

The leading AI has already forked into two options. A: Closed source American that costs $26-56 per 1MM tokens. B: Open weight Chinese that costs $0.50-1 per 1MM tokens. If you force American companies to spend 50-100x more than their competitors abroad here is what will happen: 1. American companies spending 50-100x will at some point become financially impaired and the stock market will crater. This is the equivalent of the US Government saying we can only buy oil at $800/barrel even when the free market sells the same oil for $80/barrel. 2. In the short term, the revenues of the Closed American Labs may remain hi. But then their revenue will also crater because their customers in the US will be impaired and their customers abroad will have already flipped to cheaper solutions. This would be a terribly self-defeating form of intervention if it were to happen.

@Dio76Bri76 @chamath Thats actually very true

@Marvinma12 @chamath Kimi is struggling with compute today. K3 would never had gotten to this point without OpenAI and Anthropic. Chinese AI has benefitted from those American expenditures.

A moment frozen in time from 2,600 years ago. Archaeologists from the Israel Antiquities Authority and Tel Aviv University have uncovered charred wooden beams in Jerusalem’s City of David, burned during the destruction of the First Temple in 586 BCE. Discovered during excavations at the Givati Parking Lot, the thick beams most likely formed part of the roofing of an inner courtyard. Researchers believe that plaster covering the building’s walls melted during the intense fire, covered the charred wood and helped preserve it for approximately 2,600 years. Sometimes, history does not just live in books. It waits beneath the ground to be rediscovered. 📸 Assaf Perry, For the City of David and the Israel Antiquities Authority.

@mark_k @OpenAI I dont think its that simple, its a benefit on one hand while also more expensive interms of context size on the other

ChatGPT news: @OpenAI just increased the custom instructions limit in ChatGPT from 1,500 to 5,000 characters for Plus, Pro, Enterprise, Business, and Education users. That is a massive upgrade for anyone who wants ChatGPT to consistently follow detailed preferences, workflows, formatting rules, and writing styles. 1,500 characters always felt far too restrictive. 5,000 is finally enough to make custom instructions genuinely useful.

@chamath The fact is one is open while the other closed

@chamath Its quite a crazy situation, I don't know why US companies are spending so much (with exception on compute). I think they need to get their act together otherwise, they risk being the first to bust in this AI era

@benln Thanks for sharing Ben

@Marvinma12 luma.com/cursor-mzjz

Cursor team will be visiting these cities over the next few weeks: • Nairobi • Bangkok • Da Nang • Tokyo • Sapporo • Accra • Bali • Kathmandu • Delhi • Bangalore • Casablanca • Mumbai • Hyderabad Sign up on Luma (limited spots)

@Reuters Wow! More money more investments. I think this bet will be worthwhile

UK government, Bezos back CuspAI's $450 million round as startup seeks to discover new materials reut.rs/4ww1AwR reut.rs/4ww1AwR

@mark_k Google is really firing on all cylinders, I wonder why they are delaying their 3.5pro, the competition is seriously overtaking them in that section.

Google is working on a very different kind of AI chip called "Frozen v2." Instead of merely accelerating Gemini, it would hardwire parts of the model directly into the silicon. The result could be 6–10× more tokens per watt than Google’s latest custom AI chips. Frozen would complement Google’s TPUs rather than replace them, with deployment reportedly targeted for 2028. This is probably where AI hardware is headed: models and chips designed together so tightly that the boundary between software and silicon starts to disappear.

@mark_k Whats loading this week, if its earnings, then a possible slide in the markets

BIG WEEK for AI coming. 📈

Scott Galloway predicts a familiar pattern in AI including disappointing applications, overbuilt infrastructure, and eventually, capital markets adjusting. He warns that an increasingly capital-intensive, competitive, and commoditized market might mean no one wins enough to justify current valuations.