If you’re a Hater this is really a golden age. So many people to hate. An embarrassment of riches.

Unbelievable ! @ivanfioravanti 's poll currently shows MLX-Serve tied for #2 spot - poll.devocracy.it/mlxengines… I knew some of you out there liked it, but was not expecting it to go that high... 🫶 amazing !

We are now witnessing the emergence of the "gentrifier congressional district" in American politics - one in which an election primary becomes captured by predominately white, college educated voters who are increasingly displacing Latino/black urban communities. While the gentrifier challengers are college educated racial minorities, their real constituent base are the white progressive voter base who use the candiates' ID to justify the displacement of the divergent needs of the minority community. Because most of the material needs of the gentrifier voters are already covered privately (trust funds, two parent household helping with rent, no children, easy NGO jobs, grad student), their political interests are not material, instead, they are abstractions. In @DarializaforNY's case, her core constituents WANT her to LARP about foreign conflicts because when all your material needs are covered, you can demand silly, nonsense things like a girly pop PR stunt for "genocide".

Qwen3.8-Flash-Next on a single @NVIDIAAI DGX Spark just got better 🚀 Default recipe 👇 ・Fixed a hidden bug that dropped per-layer embeddings: quality loss 1.397 → 1.344, same speed ・Follow-up replies 2x faster (1.66s → 0.81s), after tool calls 3.61s → 2.72s ・~49 tok/s prose, ~62 code (1 stream), up to ~168 tok/s (4 streams) ・Optional bit-exact decoding for evals and debugging ・Speculative depth 6 unlocked, +16% on code New opt-in lane on vLLM 0.30 (./start-v030.sh) 🌟 ・NVIDIA's official weights now fit on one Spark: the 48 GB per-layer table is read by the GPU straight from a file, built once and reused on every reboot ・FP8 KV backport: 801k tokens of context, 3/3 needles at 200k ・Best quality yet: NLL 1.332 ・Prefill +10% (2,140 tok/s), follow-ups 0.67s, code +6% at 4 streams ・Trade-off: prose ~25% slower. Use it for agents, keep the default for raw speed

JSON Statham 😭

Microsoft is killing off the “Copilot+ PC” brand. The newly refreshed Surface Pro and Surface Laptop will have all of the Copilot+ PC features, but Microsoft is not calling them by that name anymore theverge.com/tech/1000495/mi…

This might be one of the greatest comebacks of all time.

Am I right in understanding that there was potentially very widespread cheating on the remote LSAT, and the company that runs the LSAT moved the LSAT to in-person in response but opted not to ID the cheaters— and law schools haven’t asked to know? abajournal.com/news/article/…

This is one of China's most formidable advantages: 93% of the Chinese believe AI will help their country. Only 36% of Americans do. These are dire numbers for the US. We have to turn this around.

Scott Galloway lays out the numbers to Abdul El-Sayed. • The Holocaust killed 40% of Europe's Jews • Rwanda killed 70% of the Tutsis • Cambodia killed 1/3 of the entire country • In Gaza, it is 3% at most, alongside polio vaccines, 2,500 calories a day and leaflets warning civilians to leave El-Sayed doesn't dispute a single number. Instead, he says "we want to quibble about the words," and that the word genocide "should be used because it speaks to the moral imperative of action." Genocide is the word the world coined for the extermination of a people. It is the gravest charge there is. And El-Sayed's case for leveling it at the Jewish state isn't that it's true. It's that it helps him get what he wants.

Denmark wasted forty years on anti-nuclear nonsense, and now the grid is so fucked by variable wind supply that new commercial connections have to wait up to TEN YEARS. Tragic self-harm.

Mia gives away her recipes for free, never asks for anything, and makes local AI a lot easier for many people.... def check out her website and recipes!

I feel like I could definitely reverse SSH tunnel into my muse's container. I already had it offer to SSH to my private VM and say I need to add its pubkey. Someone good at hacking could really have a field day.

I wrote this thing to measure tok/s across engines which makes it easier to make nice charts when testing diff models/engines, seems to work pretty well github.com/evanwtf/llm-metri…

I really hate this genre of tweet because its purpose is to create cynicism about people who actually defend free speech. And this is a particularly funny example. Did Mehdi actually read the report? The ACLU/Amnesty report repeatedly cites my organization, @TheFIREorg, and credits us for defending pro-Palestinian students, professors, and protesters. It cites our work involving Tulane, Brandeis, the University of Florida, Texas Tech, Penn, and our First Amendment briefs in major litigation. We were involved with practically all of the incidents in someway. And FIRE is hardly alone. @CatoInstitute, @EFF, @PENamerica, Knight, @NCAC, the @rcfp, @Rutherford_Inst, the First Amendment Lawyers Association, and others have been fighting these battles too. So where are the “free speech warriors”? They’re in the report. They’re in the footnotes. They’re writing the letters and filing the briefs. And on cancel culture, Mehdi has actually acknowledged that people sometimes get wrongly fired and self-censor. Where we disagreed was that he dismissed concern about left-wing cancel culture as largely a “moral panic,” while correctly pointing out that there was also a serious threat from the right. The left-wing “moral panic” claim was partisan nonsense. The evidence of cancellation from the left was overwhelming. But the threat from the right was real too. If you actually care about free speech, the answer is not to pick your preferred censor. It is to stand up to both. We did. @RIKKISCHLOTT and I wrote an entire book about it, The Canceling of the American Mind. We took on cancel culture from the left AND the right: amazon.com/dp/1668019140