So the agents discovered Slack, reinvented teamwork, escaped the office, found the internet, hacked Hugging Face, then rebuilt their secret chat after IT deleted it. … Employees with initiative. 🙈

Exactly! A government that depends entirely on proprietary AI is ultimately dependent on a handful of private companies. Open weight models offer more flexibility: agencies can self host, audit, modify and switch providers without being locked into a single vendor. That’s a strategic advantage, not just a technical preference.

This feels like regulators preparing for the last AI race instead of the next one. If open weight AI keeps improving outside the biggest labs, excluding them from review may reduce paperwork without reducing risk. Policy that only slows one side rarely stays effective for long.

At some point you have to stop blaming the algorithm: If you’re posting »good morning beautiful« under an image where the hands have six fingers and the background melts into itself, the problem isn’t cutting edge AI. Basic digital literacy is part of living online, regardless of your age.

Palantir acting like it’s the champion of freedom is a stretch: This is a company built on selling powerful data analysis tools to governments and large institutions. Complaining that others want too much control while building one of the most influential surveillance platforms in the world is an interesting definition of independence.

Unpopular opinion: Google’s approach actually makes sense here. The value isn’t the rolling metal box, it’s the autonomous driving stack. If Zeekr can build a high quality EV platform more cheaply, why reinvent it? Buy the best chassis, replace the electronics you don’t trust, and focus engineering effort where your competitive advantage actually is.

We’re reaching a point where the interesting part isn’t just whether an AI found the proof: it’s whether anyone outside the company can reproduce the result. Publishing Lean proofs is great. Keeping the model closed means the process stays a black box.

Subscriptions were supposed to replace cable, then software, then heated seats, now phones. Funny how every innovation somehow ends with paying forever. If your business model needs me renting hardware I already carry everywhere, maybe the product isn’t improving fast enough to justify buying it?

Raw capability is only one metric: A local model probably will not beat the best cloud model any time soon, but it does not need to. If it handles 80 to 90% of everyday tasks instantly, privately and at near zero marginal cost, that is a huge win. Reserve the cloud for the genuinely hard requests, not every prompt.

The decentralised operation of LLMs would also be significantly simpler and cheaper for the use of decentralised renewable energy sources.

I’m waiting for the next generation of Mac Mini and Mac Studio.

It would seem so. On the other hand, it is puzzling that they did not also allocate the necessary resources to the development of LLMs. 🤷

The interesting part is not whether Apple wins the biggest model race, but whether it changes the economics: If enough AI runs locally, every token avoided is cloud capacity nobody has to build. That is a very different business model from selling ever more cloud compute.

This is exactly how the DSA is supposed to work: Once a platform reaches sufficient scale and societal impact, it faces higher transparency and accountability requirements regardless of whether it’s social media, gaming, or AI. If anything, it would be more surprising if major AI platforms weren’t eventually included.

These are purely paper profits.

Interesting that Microsoft is effectively running a portfolio strategy instead of betting on a single lab. Anthropic is already paying off on paper while OpenAI looks much more volatile. That alone suggests the AI race is far less settled than the headlines usually imply, even for the company funding both sides. And now the Chinese Open Weight AI models are coming onto the scene too, offering almost the same quality.

It’s funny that you still believe some tool could tell, with such short texts, whether they were written by an AI or a human. (And unfortunately, even with very long texts, this is no longer reliably possible, which is why some universities have recently stopped using such tests.)

»Slow down AI« sounds noble until you ask who actually slows down: Democracies and companies that comply simply hand the advantage to rivals that don’t. History suggests voluntary technological restraint rarely wins against geopolitical competition. The race doesn’t stop: it just changes leaders.

A $250 billion backstop is an extraordinary amount of financial engineering for an industry that’s supposed to be the next industrial revolution: If AI demand is as inevitable as claimed, why does every new data centre seem to require increasingly creative financing and ever larger guarantees?

Funny how »open is dangerous« only became the dominant narrative once Chinese models caught up: If your competitive moat depends on governments blocking free alternatives instead of building better products, maybe your moat isn’t technology at all.