buckeyeVR

@frozenpepper Are you using anti-gravity or the CLI? Is it safe to say that using the flash model to do the initial programming and then follow up with a higher level reasoning, model find tune the output is the way to go?

Been coding, HARD, with brand new Google Gemini Flash model (3.2 or 3.5 however they are going to call it) for 24 h straight now. Some impressions : - Lightyears better than previous flash model - Fast, like scaring fast, like so fast it makes you doubt it actually did the job 🧵

@AnthemBCBS And once again, fuck @AnthemBCBS !

@AnthemBCBS I appreciate you reaching out, but I’m really not sure what good it’s going to do! The idiot doctor on your end is not allowing the surgery to happen even though her surgeon says it’s necessary!

@AnthemBCBS I’m not sure what to say other than your insurance company is terrible! My wife has been in back pain for two years! She has had multiple injections multiple diagnostic imaging showing narrowing of the disc spaces and you won’t approve the surgery! Go fuck yourself!

@Quadro_News Any x6 news?

👉Well, then. Today, DJI has some good news for us: the company has officially announced the launch dates for the DJI Mic Mini 2 and the DJI Lito. In addition, we get a first look at the new design of the Insta360 Go Ultra.

@itechnologynet @TeksEdge @grok @grok how can I use both of these on a Mac?

@TeksEdge @grok compare this to rotorquant

📍 Local Inferencing News: TriAttention Wins over TurboQuant for Local Long-Reasoning. 🚀 Both tackle the KV cache bottleneck in LLMs but TriAttention is superior for consumer hardware & long reasoning. Here's the breakdown: 📊 Memory Savings ✅ TriAttention: 10.7× less KV cache (prunes low-importance tokens entirely via trigonometric scoring in pre-RoPE space) → Caps at ~0.82 GB even at 60K+ tokens ✅ TurboQuant: ~6× reduction (3-bit keys / 2-bit values quantization) Winner: TriAttention where true sparsity beats bit compression for extreme contexts ⚡ Speed & Throughput ✅ TriAttention: 2.5× higher inference speed on long reasoning (e.g. AIME25) → Decode stays ~10 t/s (stable even at 60K tokens) ✅ TurboQuant: Up to 8× faster attention logits (on H100 GPUs) Winner: TriAttention has real-world end-to-end gains on RTX 4090-class hardware 🎯 Accuracy & Use Cases ✅ TriAttention: Zero accuracy loss on reasoning/math benchmarks (perfect match vs full attention) → Built for generative long-context tasks (reasoning, code) ✅ TurboQuant: Zero loss claimed (training-free) but community tests show slight degradation vs traditional quants in some cases Winner: TriAttention uses trigonometric pruning preserves patterns better for reasoning 💻 Hardware Fit ✅ TriAttention: Runs 32B models on a single 24GB RTX 4090 (vLLM-ready, 1-click) ✅ TurboQuant: Strong on datacenter GPUs; less optimized for consumer VRAM limits Winner: TriAttention enables massive local models Bottom line: For local 32B+ long-reasoning on a 4090, TriAttention is the clear winner right now. Pruning > pure quantization for speed, memory, and accuracy in real generative workloads.

@jtdavies How are you running the turboquant

MLX + TurboQuant = Local Super Power Take a local (private) document(s) or codebase, pre-fill the 256k KV cache (the context) with the document(s) and system prompt, quantise and run on Apple's MLX and you have almost instantaneous, lossless document queries with total privacy. For a 75-page PDF (some 30k tokens), I was seeing sub-150ms TTFT (Time To First Token), that's "instant" for a human. We run this on an office server (one of our Mac Minis) for staff to work with, instant answers on critical, confidential company/client documents on phone, iPads or laptops. Imagine the "What were the 3Q25 profits?", "When did Joe start working for us?", "Please list the project milestones and team count" answered within seconds of hitting the return key. It works with voice too. This example is 17k lines of code from 177 source files plus an index of the entire 1,902 files, the entire Claude Code codebase (oops Anthropic!). Zero loss (unlike Anthropic's security). What this means is the ability to work better on local (private) code bases, off-the-shelf private document analysis with better response than the best public models - all locally.

@elonmusk @grok how do we access the coding models

By this weekend, xAI will have three Grok Build models in training simultaneously

@vikaskansalHQ How about the way your company handled this was absolutely a clinic on how not to do things! I don’t think any of us had any notifications or emails sent about the upcoming change, which would’ve been a simple task. I don’t know that I would’ve paid for a full year had I known

We are excited to expand AI Credits, our currency for compute to Antigravity to enable the next generational use cases foe agentic workflows. Please share your feedback on what we can do better!

@antigravity Shit move all around!

We’re evolving Google AI plans to give you more control over how you build. Every subscription includes built-in AI credits, which can now be used for Antigravity, giving you a seamless path to scale. Google AI Pro is the home for the practical builder, hobbyists, students, and developers who live in the IDE and don't necessarily rely on an agent. This plan features generous limits for Gemini Flash, with a baseline quota included to "taste test" our most advanced premium models. Google AI Ultra serves as the daily driver for those shipping at the highest scale who need consistent, high-volume access to our most complex models. If you’re on Pro but need "extra juice" for a heavy sprint or deeper access to premium models, simply top up your AI credits to customize your plan. Keep building. Keep shipping.

@AbudBakri How effective do you think it would be for narcolepsy or idiopathic hypersomnia?

Took a few weeks off pinelaon and then restarted lol I still can’t believe it every time. If this is what baseline is supposed to be, why aren’t we there?

@BasedBiohacker @FitFastCoach How common are sleep issues with modafinil?

@FitFastCoach it's incredible. did you have sleep issues?

it's your very first time taking modafinil? congratulations. i'm proud and excited for you. if i were you, this is exactly what i'd do; 1. assess your stack - remove all stimulants and keep only things like vitamins, herbals, and minerals. i.e., only things that do not provide immediate efficacy. 2. start off light. 50 mg should do it. take it on an empty stomach early in the morning. take it with 300 mg alpha-gpc to avoid choline deficiency. 3. assess efficacy and feeling during the day and at the end of the day in detailed writing. 4. repeat and dose up to 200 mg in 50 mg increments. so 50 mg → 100 mg → 150 mg → 200 mg. 4. assess your notes and impressions, then conclude on your ideal dose. 5. slowly start introducing new compounds with immediate efficacy if desired. repeat the earlier process. 6. congratulations - you have built your ideal stack. now, get to work on achieving those dreams. systematic & scientific > winging it

@JBook_37 @DylanEveryday We need to get that in the urban dictionary!

@DylanEveryday Allegedly pulled a Sheronne

Does anyone know what he did

@Genetics56 The nine most terrifying words in the English language are: I'm from the Government, and I'm here to help. -RR

Ask the government to come into your home and you will regret it. Many folks, including myself, said it.

@KurtisCarman @SkiveWalker @JBook_37 Almost night and day!

@SkiveWalker @JBook_37 It was unbelievable how much better the whole thing was when he was out.

Tegra might grade out with a negative score when PFF releases their grades tomorrow. That boy got Molly Whopped. The entire OL got taken behind the woodshed today. I will be surprised if anyone on the OL grades out above a 65%.

@JBook_37 Just go buy the best in America but the coaching staff is stubborn and won’t do it!

Jayden Fielding has cost Ohio State some huge games not being able to make kicks inside the 25 yard line.

@LandonTengwall Everyone but the coaching staff sees it

Ohio State is going to continue to have problems at RG until they stop playing Tshabola

@JBook_37 This is just awful…

My god how is that not a hold again?

@perplexity_ai something happened to my free faculty educator account and I’ve had no reply from your team on how to get it back? Any thoughts on what I can do?

@PSUBarstool Everyone forgets the first couple years at OSU and just remembers last season. Look back at some of his disasters and you will see the full picture.

What does Jim Knowles even do?