HallidayZero

@gridpane @garrytan Agreed Qwen 3.6 is awesome. And if you want decent throughput, you might as well go with Qwen3.6 35B A3B which performs really well for its size profile.

@garrytan I doubt this is gonna perfect better than Qwen3.6 27B dense. I just don’t see it beating that at 2bit quantization. The million token context window is also kinda bullshit because they all fall off the earth at half their full cap. But it’s where the puck is going!

Downloading now... 1M token context window with supposedly usable coding agent capability all on a 128GB Macbook Pro is 🤯

@ThePrimeagen @MatthewBerman You get insurance for uncertainty. You unknowingly described the bull case for this startup.

@MatthewBerman Insurance requires statistically understood risks such that you charge X and understand over a broad distribution you pay Y of which X = Y + c where c is your profit We have no clue what 2027 will hold and the benefits or dangers of AI

This has to be a scam.

@MoonsAndLambos @coookzy Lore accurate kid

@coookzy so cool we have the same mom!

This is my Mom, she works at JPMorgan Chase and has recently been SLANDERED in the news by a disgruntled ex-employee. My mother, Lorna, is loving and has always put family first and taught me the importance of respect, hard work, and standing up for what’s right.

@MVetisan @postdiscipline @ReddCinema @UniversityFL @grok It’s not a problem. That’s probably why those schools are so good. It is however quite dishonest to say UC schools discriminate against Asians when they are the majority and over represented by 8-9x.

@HallidayZero @postdiscipline @ReddCinema @UniversityFL @grok verify this statement And even would be true what is the problem if they are the best?

He had a GPA score of 97.3%, SAT score is 1560, enrolled in a top high school, and does lots of extracurricular work. He got rejected by multiple colleges. They call this equity and inclusion.

@Evaebonyy @saint_cloudy Yes

@saint_cloudy All passengers have 90 seconds to leave the aircraft during an emergency evacuation… so you’d rather be trying to reach for laptop bag than save your life?

Wallahi I’m taking my laptop bag with me you niggas are crazy

@MVetisan @postdiscipline @ReddCinema @UniversityFL The UC schools are like 60% Asian. What are you even talking about?

He got rejected by UCLA, Berkeley; accepted into UC Riverside, UCSB, which he doesn't care about. Re application, the 9 UCs use ApplyUC portal where AP scores get entered into preset fields. The top UCs clearly r discriminatory towards white boys and Asians from what I gather. We live in SD and I am pissed for being rejected by his choice, UCSD, while we pay insane $$$ taxes, forgetting about his grades which r stellar. F California.

@SanDiegoKnight TLDR, five ways to cope with immigrant success

This Cato ‘immigrants pay more taxes’ flex + Indian chart is peak cherry-picking. Impressive numbers… until you actually look under the hood. 🧐” 1. It’s median household income, not individual or per-capita — and Indian households are structured differently • The chart (and the “twice as much” claim) uses households, not people. Indian-American households are larger on average (~3.0–3.8 people vs. U.S. average ~2.5) and far more likely to have multiple full-time high earners (dual STEM/medical professionals is common). en.wikipedia. • Indian Americans still have high personal earnings (median ~$85k for ages 16+, ~$106k for full-time workers per 2023 Pew), but the “almost twice” headline evaporates when you adjust for household size and number of workers. This is a classic statistical sleight-of-hand when comparing groups with different living arrangements. 2. Extreme positive selection bias … this is the cream of India’s elite, not “immigrants” in general • Indian Americans aren’t a random sample of India’s 1.4 billion people. The vast majority arrived via H-1B, EB-2/3, or student visas …hyper-selective for advanced degrees and high-skill jobs. You’re comparing the top ~0.1–1% of India’s talent/IQ/education distribution to the broad U.S. average (which includes everyone from McDonald’s workers to retirees). • India’s own per-capita income and education levels are far lower. This doesn’t prove broad immigration is economically magical; it proves cherry-picked high-skill immigration works for the selectees. Second-generation outcomes are strong but show some regression toward the mean, and chain migration/family sponsorship often dilutes the skill level over time. 3. Cato’s overall “immigrants pay more taxes” claim has well-documented methodological holes • Cato (a libertarian think tank that favors more immigration) attributes welfare benefits received by U.S.-born children of immigrants to “natives,” not the immigrant parents. This understates immigrant fiscal costs. The Center for Immigration Studies (CIS) and others note this flips the picture: when you count the full household burden (including kids), immigrant-headed households use welfare at higher rates than native-headed ones. • Cato aggregates all immigrants (high-skill Indians + low-skill groups + illegals). The net positive they find is heavily driven by the high earners. Other studies (National Academies of Sciences, Heritage, etc.) have found first-generation immigrants often impose net costs, especially low-skilled/illegal cohorts. • Their data ends before the post-2021 border surge effects fully hit long-term budgets. 4. H-1B-specific issues (the main pipeline for Indian success) • Many Indian immigrants in tech come via H-1B, which has documented problems: outsourcing/body shops (e.g., Infosys, TCS), wage suppression (foreign workers often paid less for similar roles), and ethnic nepotism once Indians reach management (preferring co-ethnics for hiring/promotions). This displaces U.S. workers and depresses wages in STEM. • Fraud allegations are common (fake credentials, benching workers, etc.). Critics argue this isn’t “adding value” so much as arbitraging cheaper labor and networks. 5. Other drains and context • Remittances: Indian Americans send massive sums back to India (India receives over $100B+ in remittances annually, a huge chunk from the U.S.). That’s money leaving the U.S. economy. • Cost of living: Indians are heavily concentrated in high-cost metros (SF Bay, NYC, etc.), where nominal incomes are inflated anyway. Adjust for purchasing power and the gap shrinks. • The post uses Indian success to defend a general “immigrants = net positive” narrative from Cato. But Indians are ~1.4% of the U.S. population and an outlier. Broad policy implications (more low-skill immigration, open borders, etc.) don’t follow from one high-performing subgroup.

@ICEgov The one job you wish Immigrants had taken

@ericcouu SF. They’re all introverted nerds. The only couples there are AF WM. And those two pair up everywhere anyway.

Is dating worse in LA or SF?

@rcc @SpencerHakimian Booty

@SpencerHakimian 🤔 I’m slow. What is it that I’m suppose to notice?

Petr Hegseth’s with is going viral.

@DavidSacks There are lies, damned lies and statistics.

Narrative violation: Hiring of new college graduates is up 5.6% over last year. Youth unemployment for degreed 20–24‑year‑olds fell to 5.3% from 8.9%. Weren’t we told that 50% of entry-level jobs were going away?

@quintasluis29 @Alibaba_Qwen @mr_r0b0t @grok Yes you can. Technically you can run it at q4 on as little as 32gb ram.

@Alibaba_Qwen @mr_r0b0t @grok can i run this on my mac studio m4 ultra with 96gb of ram?

🚀 Meet Qwen3.6-27B, our latest dense, open-source model, packing flagship-level coding power! Yes, 27B, and Qwen3.6-27B punches way above its weight. 👇 What's new: 🧠 Outstanding agentic coding — surpasses Qwen3.5-397B-A17B across all major coding benchmarks 💡 Strong reasoning across text & multimodal tasks 🔄 Supports thinking & non-thinking modes ✅ Apache 2.0 — fully open, fully yours Smaller model. Bigger results. Community's favorite. ❤️ We can't wait to see what you build with Qwen3.6-27B! 👀 🔗👇 Blog: qwen.ai/blog?id=qwen3.… Qwen Studio: chat.qwen.ai/?models=qwen3.… Github: github.com/QwenLM/Qwen3.6 Hugging Face: huggingface.co/Qwen/Qwen3.6-2… huggingface.co/Qwen/Qwen3.6-2… ModelScope: modelscope.cn/models/Qwen/Qw… modelscope.cn/models/Qwen/Qw…

@garrytan You've lost the Plot buddy. We love you, but get help.

Now launching GStack Browse Before, /browse actions are by default headless, but now you can just ask for headed mode and you get a real steerable browser. The sidebar is an interactive Claude Code session that lets you navigate, run operations, and is an open source customizable version of what Comet or Atlas Browsers give you It's connected to both the sidebar AND your origin Claude Code instance so it's useful for things like page debugging and CSS interaction Try it now: run /gstack-upgrade and then /open-gstack-browser

@thejuggernaut Daddy Issues Final Boss

Priya Patel went viral for her take on immigration. Then, one interview took an unexpected turn.

@TPA_Patriot @jessegenet @GoogleDeepMind @openclaw Those figures are significantly lower than they should be. Are you using MLX? It can be > 50% faster.

@jessegenet @GoogleDeepMind @openclaw Hey Jessie! We just received our 512 gb ram studio for our work. I downloaded all 4 - Gemma 4:31b Qwen 3:235b Qwen 3.5:122b Qwen 3.5:35b And said “run a benchmark based on what we do/have set up at our company” and this is what it found. I would have your agent do the same!

It’s happened. Mac Studio is here. Gemma 4 31b @GoogleDeepMind installed, chatting with my main @openclaw for $0 in token expenses now... I've burned $5-6k on tokens on my crazy ideas over past few months, so this mac studio should pencil out for me within 3 months or so 🤓

@RickEllisonNY @mattvanswol @catturd2 Genuinely cant tell if you’re being sarcastic or serious

@mattvanswol @catturd2 ICE checking everyone at airports is wonderful. Here are more great ideas. 1. Traffic courts. Check every person who comes into the building. 2. DMV offices. 3. Welfare department offices 4. Bus and subway cars. Board with a team and check every single person onboard.

Dear ICE …. While you’re there, arrest and deport every illegal going through the airports.

@__tinygrad__ So your grades in a college class put you on the same level as Jeff Dean?

Few know this, but I (George) was the only person in history to get a perfect score in CMU compilers, which is likely the best compilers course in the world. Combine that with crazy low level knowledge of hardware from 10 years of hacking. Then add a team of people who are talented enough to push back on my dumb ideas and clean up the implementations of the good ones. The team who keeps this whole operation running, software, infrastructure, and product. I love how there's no hype in deep learning compilers. It was one of the most annoying things about self driving cars, all the noobs who burned through billions on crap that was obviously dumb, and the companies who deserved to go bankrupt years ago if not for government bailouts (Tesla and China will devour them all). In this space, the competition is @jimkxa at Tenstorrent, @clattner_llvm at Modular, and @JeffDean at Google. Three of the living legends of computer science. And companies like @nvidia and @AMD, who are definitely live players, making single chips that have more power than the whole Internet two decades ago. This space is so fun to play in. If you haven't, read the tinygrad spec. It's all coming together beautifully.

@n0w00j @adisingh

@adisingh FUCK YOUUUUUUUUUUU

im sorry but i just don't get how this YC company is a viable product when this API exists. maybe i'll launch Powerpoint for Agents

@chribjel “But hE hAs dIstRibUtIoN”

theo, the king of wrappers t3chat -> llm wrapper t3code -> agent wrapper whats next?

@geoffreywoo @sneezuschrist @harvey @RogoAI @OpenAI @AnthropicAI Wrong, it just means you need tuning pipelines fast enough to keep up. The only edge you’re going to have over your competitors is leveraging your proprietary data.

@sneezuschrist @harvey @RogoAI @OpenAI @AnthropicAI no point tuning models if openai, anthropic, and gemini are releasing new models every 1-3 months

short: @harvey @RogoAI (legal ai and finance ai) long: @OpenAI @AnthropicAI models outpace any RL/fine tuning, and then vibe the UI/UX that wrapper cos tout invest in foundation labs, and only startups with proprietary data moats. models on pace to swallow everything