draxen
I wonder what the biggest dense model in the same model family, the surprisingly capable Qwen3.5-27B, would be able to do if given the same task.
For the rest of us, that also means the S&P 500 free lunch* will soon be over.
American zillennials are going to get their free healthcare and free higher education, and they'll do it in a way that makes France look like a well-managed, investor-friendly country.
*Unless you own real estate in Copenhagen, one of the biggest sources of wealth creation in Denmark since 2015 has been our pension funds and their exposure to US stocks.
>2. Would an equivalent allocation to incentivize human effort have borne the same results?
Unlike their initial massive brute-force run that solved the Navier–Stokes Millennium Prize Problem, most of their subsequent breakthrough results in mathematics have come from a single prompt to a single AI agent, using approximately 3 hours of ChatGPT Pro-equivalent thinking compute per result on average.
I do things that require ~300 tool calls per prompt and I have been stuck with OpenAI and Anthropic since 2024.
If people still wonder what moat OpenAI and Anthropic have: They got 100% tool-calling accuracy (using constrained decoding) and prompt caching right all the way back in 2024.
The word is absolutely meaningless now.
they have never been able to keep up with Groq/Cerebras-level inference speeds for some reason.
Unfortunately the distance between stupid, self-defeating Anglo-American activism and European legislation is shorter than the distance between that activism and American legislation.
Generally, Europeans would benefit from being as critical of progressive Anglo-American activists, academics, and media as we are of American corporate interests.
The raw problem solving capabilities of Chinese open weights models are still weaker than or on par with Opus 4.5 but they have been post-trained to be much better at agentic coding, cybersec, and computer use.
And that is okay. GLM-5.x and DeepSeekV4.x have both been highly useful on their own.
Together.ai, Baseten, and others have already achieved higher throughput, lower latency, and lower E2E latency for DeepSeek V4.1 Flash 1017 than DeepSeek's own inference service.
Violating the E (Environment) should lead to fines to make up for environmental clean up. If severe enough there should be prison time for the responsible higher ups.
Drop the S and G. There are existing laws to for most of these.
Governments can use weapons embargoes to stop their defense contractors from selling to specific countries.
The SFDR and similar EU regulations were originally designed to tackle greenwashing, but has now made it almost impossible to raise money for nuclear, aerospace, and defense startups and ventures in the EU (and by network effects also in the UK).
Since 2025/2026 the EU has promised to find ways ease up these regulations because of the war in Ukraine.
I remember NYTimes/WaPost pieces being concerned about Hindu nationalist influence on the Republican party. Anyone who has spend 2 seconds online know it's more likely the Republican party nukes India than forming alliances with the BJP.
One problem the Brazilian left is facing is that the non-Musk platforms are not too fond of death threats against Protestants, landlords, farmers etc.
Which is weird. CDN requires massive distributed footprint, which inherently provides the enormous network capacity needed to absorb DDoS attacks. So why aren't they offering the same thing as Cloudflare?
European CDNs should form partnerships with both European IXPs and Tier-1/Tier-2 ISPs (like Orange, Deutsche Telekom, etc.) to see what exactly is missing to do both unmetered anti-DDoS and global CDN/caching.