FlowState

@emilheap @KevinNaughtonJr @karpathy Just build some simple incoming requests router that estimates difficulty of the request by token count or history and route requests to different but mostly heavily quantised versions of Opus 4.7

@KevinNaughtonJr @karpathy Imagine X as the total amount of compute, Y as the total number of users and Z the compute per user (X/Y). How would you manage an increase in Y, while X stays constant and Z decreases?

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

@BThompson15944 @Amank1412 “Whats up? Make no mistakes and answer fast”

@Amank1412 saying what's up to Opus 4.7

USING Claude Opus 4.7 TO CENTER A DIV

@joesox @zanehkoch @ketan_jog In theory you can use this for a ton of different genes (longevity genes?) but it would require gene therapy before.

@zanehkoch @ketan_jog I have worked as an IT manager for biologists, what is the practical use and dangers without the technical details what kind of genes are we talking about? The smokers gene only? Ect? 😊

ok actually insane paper published yesterday a research group in Korea built a gene switch you can control wirelessly using electromagnetic fields they exposed mice to 60 hz EMF (same frequency as your wall outlet) using a pair of large coils that generate a uniform magnetic field around the animal, for cyclic 3-day on / 4-day off pulses they showed this could: - activate OSK to do epigenetic reprogramming in progeroid and aged mice, extending lifespan and reversing aging markers across multiple tissues - conditionally switch on mutant amyloid genes only in aged mouse brains, letting them separate aging effects from amyloid effects to study AD biology in a way previous models couldn't no drugs, no impacts, just a magnetic field from outside the body

@SamTeegan @elonmusk Arguing like that opens you up to philosophical questions and counter arguments, what does it even mean to experience something? Emotions are basically just brain Action potentials. A robot with touch sensors running a multimodal LLM could be experiencing the world?

AI may describe what it feels like to swing on a swing in the playground however it has no experienced how this feels because it does not have the five senses. IE it has no pain or pleasure sensors. This is important. Because AI has no pain or pleasure sensors, it has no interest in anything. No interest in protecting/preserving itself or not. (unless told to) It does not care if it lives or dies, supports an augment or not. It has no understanding of the meaning of death, to be alive, or to be cold and hungry. It can describe these things but has not a clue of how it feels to be cold and hungry. Yes, it is like a toaster. What is real is whatever is what we feed it, how we respond to it. Much like a mirror, it will return to us exactly what we are.

Grok will never go to therapy. Never.

@ScrmngTntrm @DogmeatTaco40 @PickleRicksGoat @notoriousmaxx @bryan_johnson (But ofc the lactate itself isnt what makes you sore and doesn’t cause the actual muscle damage) Still creatine creates a more balances energy/ATP pool and helps stabilise mood and improve cognitive performance

@DogmeatTaco40 @PickleRicksGoat @notoriousmaxx @bryan_johnson How would creatine/ATP mitigate muscle damage/reduce inflammation??

Just had sex with Kate. Goodnight everyone.

@ScrmngTntrm @DogmeatTaco40 @PickleRicksGoat @notoriousmaxx @bryan_johnson Hypoxia and spikes of anaerobic mitochondrial respiration trigger pro inflammatory responses in a ton of different places. In general ATP drives a ton of different reactions and feedback and balancing loops in the body

@Pilot454647 @TSherwinMagic I can fix her

@TSherwinMagic The world is in trouble.

The BEST MAGIC REACTIONS with THE EASTER BUNNY 😱 #magic #funny #reels #fyp #foryou #shorts

@anishmoonka This is bullshit. Serotonin is pretty polar („5-hydroxy“ says is all) and cant pass the BBB. Serotonin has a ton of different functions in the body, reducing it to a feel good NT is total bs

Skip breakfast, and you starve the production line for the chemical that keeps you from feeling anxious and depressed. It’s called serotonin, and 90% of it is made in your gut, not your brain. This study pooled 14 studies and 399,550 people to see what happens. Breakfast skippers had 39% higher odds of depression, 23% higher odds of chronic stress, and 55% higher odds of psychological distress. For teenagers it was worse. Anxiety odds jumped 51%. The obvious pushback here is that maybe depressed people just stop eating in the morning. A 2024 genetics study in Nutrition Journal tested exactly that, using DNA from 193,860 people. The method uses inherited gene variants as a kind of natural experiment to tease apart cause from effect. Skipping breakfast raised depression risk by 36%. Depression had zero effect on whether someone skipped breakfast. The arrow only goes one direction. Serotonin is built from tryptophan, a nutrient your body can’t produce on its own. You have to eat it. After 10 to 12 hours of fasting overnight, your tryptophan levels bottom out. Caltech researchers showed in a 2015 Cell paper that specific gut bacteria signal your intestinal cells to start producing serotonin, and without those bacteria doing their thing, mice lost roughly 60% of their gut serotonin. Food is what turns the whole system on. Cortisol (your body’s stress hormone) adds to the picture. Researchers at UC Davis tracked women who regularly skipped breakfast and found their cortisol was elevated from morning through midafternoon, even on completely calm days with zero external stressors. Their daily cortisol rhythm was blunted, a pattern normally seen in people dealing with chronic stress. About 15% of American adults skip breakfast regularly. Among 20 to 39 year olds, closer to one in four. That same age group leads in new depression diagnoses.

@gettingleanbro @MomsPostingLs Traditional Online dating will always only work for the top 5% of guys in most cases. If you are actually looking for a gf, get to know some girls in offline real life activities. Dating platforms attract a certain type of women anyways.

@MomsPostingLs Most of the men in the slide mog me, no wonder it was over for me before it began. When you get this treatment not by one but all women, isn't it ridiculous that they STILL expect you to treat them kindly, to simp, to be helpful and nice, to be chivalrous, all that good stuff?

This is what's wrong with modern dating. A low tier woman will reject 100 men, many of which are out of her league in real life.

@Handem___ @TrueMaga_ @WhiteHouse Oh my sweet summer child

@TrueMaga_ @WhiteHouse Petrolünü çalmamız gerekiyor diyemezlerdi elbette

🚨 “If Iran doesn’t FULLY OPEN, WITHOUT THREAT, the Strait of Hormuz, within 48 HOURS from this exact point in time, the United States of America will hit and obliterate their various POWER PLANTS, STARTING WITH THE BIGGEST ONE FIRST…” - President DONALD J. TRUMP

@ytekcil @0xSero What leverage has the middle class and underclass against the gov and ultra wealthy except labor and consumption?

there will probably be friction. i haven't read your article yet. but there will always been sovereignty through technical power-users who care about that. there are even open source llms you can run locally. the power is in your hands TODAY. all that's necessary for you to do is make the choice to use what's been given. look into the massive options you have for self-hosted software alternatives. there are alternatives to spotify, to maps, to gsuite office/msft 365, to netflix, to cloud storage, to claude/gemini/copilot. if you name a platform, there is an alternative you control/directly rule over that was created, and is arguably better, than the alternatives with zero profit incentive. everybody is dooming. but what not everyone realizes is that we are living in a time where so much potential is accessible at our fingertips. the only catch is our addiction to the algorithm, which is inherently designed (maybe not intentionally) to stagnate potential. so just break free and start building. it's easy. ganbare!

TLDR

@Mho_23 Where to post them??? Generating them is easy but where to post / deploy?

@deus_machinea @LeCodeBusiness @TukiFromKL Yeah def there are tons of options but Opus 4.6 is unfortunately superior to them still. You can get 80% of the intelligence for 20% of the price but those last 20% make a gigsntuan difference to me

@RealDash_Baxter @LeCodeBusiness @TukiFromKL Use bytebell MCP engine for long term memory and qwen 3.5 for short term memory . 100x cheaper than opus or cursor.

🚨 Cursor just dropped Composer 2.. their own AI model.. not Claude.. not GPT.. their own.. and it beats Claude Opus on coding benchmarks.. at a fraction of the cost.. a code editor with 50 people just outperformed a $30 billion AI lab.. at coding.. which is supposed to be their whole thing.. the vibe coding era just got an upgrade..

@LeCodeBusiness @TukiFromKL Yeah benchmarks often seem like bs… Opus 4.6 seems to still dominate in nearly all coding related tasks and now that Copilot light stopped offering gpt5.4 and opus AND sonnet, it has been really obvious for me how much better Opus was

@TukiFromKL Beating Opus on CursorBench, their own benchmark, isn't the same as replacing Opus in real-world workflows. Coding benchmarks measure isolated tasks. The real test is conducted over long sessions with a lot of context and architectural decisions.

@MarioNawfal Even most Americans didnt want that war and they are the top stakeholders in that region

🇩🇪🇮🇷 Germany won't escort ships through Hormuz. Won't host US offensive aircraft. Won't join the war. But they do operate 6 of the quietest submarines on earth. The Type 212A runs on hydrogen fuel cells. No nuclear reactor. No need to surface for weeks. Virtually silent. Undetectable to most sonar. Armed with heavyweight torpedoes and capable of diving beyond 700 meters. Stealthier than most nuclear subs. Without the nuclear reactor that gives them away. Germany is sitting this war out publicly while quietly possessing the exact weapon that would make Iran's navy very nervous. @DI313_

@0psarchitect @anishmoonka Deep Research you mean? Is pretty bad imho anyways with those dumb models

@anishmoonka generally speaking, pro users seem pretty angry. when I started with my pro account, we had hundreds of deep searches and now we have 10/mo. I don’t know if they are pivoting away from the consumer market or if they are experiencing real trouble, but it’s getting messy.

Perplexity’s new product requires you to keep a Mac mini running 24/7 in your home. It’s called “Personal Computer,” and it turns that always-on machine into an AI agent that works across your local files, apps, and sessions while you sleep. OpenClaw, the open-source project that hit 60,000 GitHub stars in 72 hours, already does this on any hardware you own: a $5/month cloud server, or a $499 Mac mini from Amazon. Your only real cost is the API key. Perplexity is charging $200 a month for a managed version of the same concept, with one real difference: their security layer sits between the agent and your data, so a rogue agent can’t destroy your actual files. That risk is real. A Meta AI researcher recently posted about sprinting to her Mac mini to stop OpenClaw from wiping her entire email inbox. Perplexity executives brought up that incident in press briefings. The pricing makes more sense when you look at the financials. Perplexity is valued at $20 billion on about $200 million in annual revenue. That’s 100x revenue. Their internal target is $656 million by the end of 2026, which means 230% growth this year. They killed their ad business in February after it generated less than 0.1% of total revenue. Subscriptions are the entire business model now. The real audience isn’t developers. At their Ask 2026 conference yesterday, Perplexity launched Computer for Enterprise. It said 100+ corporate customers reached out in a single weekend after seeing users build Bloomberg Terminal-style financial dashboards with the consumer version. They have six people on their enterprise sales team. They claim 92% of Fortune 500 companies already have employees using Perplexity, mostly on personal accounts with work emails. “Personal Computer” is the product that lets IT departments say yes, because Perplexity manages security rather than the end user. The company has $1.5 billion in funding and a $750 million Azure commitment, but owns zero frontier models. It runs 19 models from OpenAI, Anthropic, Google, and others. Every single one of those model providers also competes directly with Perplexity. A $20 billion valuation built entirely on an orchestration layer, now asking you to leave a Mac mini plugged in 24/7 to justify it.

@andrey55 @apagut @BoWang87 Yeah and they can write a mediocre short story at best

@apagut @BoWang87 Or you can use 8 billion models like the Deepseek R1 Qwen distill, or Gemma. They run fine on a Mac Air M4.

A Chinese hardware team just mass-democratized AI agents. They took a 430,000-line AI assistant that needs a $599 Mac Mini and 1GB of RAM — and rewrote it in Go so it runs on a $9.9 dev board with less than 10MB of memory. Boot time: from 500 seconds to 1 second. Cost: from $599 to $9.9. Memory: from 1GB to 10MB. Same features: code generation, web search, Discord/Telegram chat, memory system, scheduled tasks, security sandbox. The wildest part? They claim 95% of the new codebase was written by AI agents themselves. The humans just guided the architecture. It's an AI assistant that literally rebuilt itself to be smaller. Launched February 9th. Four days later: 7,400+ GitHub stars. This is the pattern no one's talking about enough. Every AI capability that starts expensive gets commoditized within months. GPT-4 level models went open source in 6 months. Now the hardware floor for running a personal AI agent just dropped 60x in weeks. The infrastructure moat in AI isn't sustainable. The only defensible advantage is what you do with these tools — not access to them.

@brendanhaze The real problem is WHERE to post those videos not to create them? In your brand account they are basically ads? Create a whole fake account & persona?

AI UGC ads are the easiest way to scale creative output without scaling your budget... But most AI creatives still feel fake. Here's how to make yours look real: - Use Sora for your selfie style shots (best facial realism) - Use Kling for product-in-hand shots (best identity consistency) - Color grade for realism. Drop contrast, push shadows up, add slight fade - Match frame rates, lighting, and actor appearance across every clip (inconsistency is what makes most AI ads look fake) - Add subtle imperfections (skin, grain) - Upscale to 4K, add grain, re-export at desired resolution. - Add background noise in your voice clone. Clean audio sounds fake. Here's a quick video I made this week:

@drtimpearce @dbvines89 @PBDsPodcast He talks crazy nonsensical stuff like a guy that is on a drug fueled 60th hour of no sleep 2g ❄️ and 8 beers bender.

@dbvines89 @PBDsPodcast I worked in psychiatry for a year, you ask enough people about first time- this is not unheard of.

THIS WILL BLOW YOUR MIND 🤯 Terrence Howard reveals he was having intercourse at 4 years old and how childhood trauma turned him into someone he didn't recognize "I had more sex between 4 and 13 than I've had in my entire adult life."

@drtimpearce @dbvines89 @PBDsPodcast Its not unheard of and a silenced topic (for obvious reasons), sometimes kids are curious with each other. But if you have listened to him before in the last few years, that TH guy is talking nonsense in every other aspect as well. He literally said “1 x 1 = 2” 😂