Gesture recognition using the Decisions API by OpenAI Developers.... Thanks to it's vision understanding it is way more accurate than my previous Jev-based demo. It was also faster. cc Tibo jason eric provencher tldraw Steve Ruizshow more

Waai Fu is adjusting the combat outfit. My last... video didn't spread that much. Apparently using more than 2 hashtags may mark a post as spam. Not to mention it was also marked as sensitive content due to "nudity".show more

Finished this one faster than I thought! The full... version is on my site right now. If you can’t support the site...it will be released here in July. Also, consider liking and sharing my vids as a free way to support my work. Having Ashley (AI Model) hulk out and level a city was toooooooo fun lol.show more

I've used ghee a fair bit in my animal-based... life but I was never able to tolerate the taste of it straight from the jar. This Great Ghee product is different. It tastes more like butter than any ghee I have ever used plus it's lab tested for glyphosate, grass fed and made from A2 cows. Ghee is clarified butter with the water and milk solids removed, leaving it almost pure fat. It is a low omega 6 concentration of saturated and monounsaturated fats and has little lactose and casein, which is why many people tolerate it better than butter and it's why it has a very high smoke point. Highly recommend this ghee and it's made by a small business in Virginia.show more

Astra (GPT-6) is here!!! I've had early access and... tested it like crazy with things like games, code, writing, browser control, presentations and general knowledge work. This is the best model I've ever used. Period. (Incredible demos below in this thread ⬇️) Here's my take on Astra: > It's insanely capable. This feels like a massive improvement, not just an incremental change. This is especially true with zero-shot prompts. > It's all about knowledge work. Slide creation, analysis, writing, and browser control. And oh my...it's so good at browser control. GPT-5.6 was already fantastic at doing things in the browser, Astra is another level and significantly faster. > We're closer than ever (arrived?) at prompt-to-playable game. And I don't just mean only playable, these are actually fun games. I bet if someone with a great eye for games used Astra, they could create a viral game within 1-2 weeks. > Astra is better at writing but not perfect. It removes much of the "AI Smell" we're all familiar with but some stink still survived. > It has a tendency to use the same design colors and look/feel as GPT-5.6 (forrest green anyone?) but it is more steerable in design than previous models. > It's highly steerable in general. A little nudge goes a long way. When I first started using Astra, almost every task I gave it would go for ~30 minutes. I wanted it to keep working. Adding more specifics to a prompt helped greatly with it's ability to work for a long time. > Astra's 3D understanding is unmatched. 3D asset creation was consistent and easy and its spacial awareness while building complex 3D worlds blew me away. I'm still getting familiar with Astra but this will now be my go-to model for any difficult work I have. Check out the demos below: 👇show more

Jev + GrokBot is the best GTM system I've... ever built It just made my GTM x200 CHEAPER and x400 FASTER than what 95% of teams are running... setup takes literally 9 minutes: prompt → GrokBot → Jev routes every candidate → GrokBot opens only survivors → campaign step 1 → open TypeSafe AI, create API key (keep it off chat paste) step 2 → tell GrokBot: store TYPESAFE_API_KEY in the secure field step 3 → prompt GrokBot: install typesafe-sdk on Agent Computer + smoke system_one (Choice) step 4 → point it at a target: "read end to end, pull positioning, ICP, every proof point. no copy yet" step 5 → add skill gtm-router: every candidate from X / LinkedIn / YouTube gets 6 typed questions → CHEAP SKIP / OPEN FULL / FLAG KOL step 6 → let Jev rank hook archetypes by viral rate, not by how often people post them - this is where the whole playbook comes from step 7 → GrokBot assembles the KOL shortlist and the 7-day plan - you stay in control of anything irreversible the result: → 3,412 candidates, X, LinkedIn and YouTube. → 15.7 seconds. $0.41. → me doing the same reading: 6h 12m. 8 grok bots brief it, Jev decides, GrokBot executes. no fine-tuning, no embeddings, no vector db - 6 typed questions and a schema. Setup your GrokBot JEV GTM today, then read the article below to learn how to build an agent team with JEV.show more

Are you safer with LIDAR, or are you safer... with vision? This is a false dichotomy. The more pertinent question today is "do you have something, or do you have nothing?" As you can see from the clips below, vision based systems avoid countless potential collisions every day. The difference between a crash and no crash isn't what sensor suite you chose — it's whether you have any AI on your car at all. Even if we concede that LIDAR may help prevent some additional crashes, we are really debating whether it is 1% of crashes or 0.00001% of crashes. Not all crashes are super complex and require lasers to detect. Most are simple, routine, and can easily be prevented by today's vision based AI. In fact, evidence is mounting that computer vision based systems can actually outperform more traditional approaches to self-driving. Why? Because the low cost of cameras enables you to create a much larger, more varied, and more diverse dataset. If you want to have expensive custom cars that's fine, but you're going to get fewer vehicles for the same budget. Seeing what's in front of you now is actually less important than predicting what's going to happen next — and the large scale datasets used to train pure vision systems are the best for predicting what's next. Counter-intuitively, the simpler and lower cost sensor actually has properties that make it better suited for training advanced AI. Computer vision based self-driving is often framed by LIDAR proponents as "cheaping out" on the sensor suite to save money. But it's not about being cheap, it's about bringing the technology to everyone. 1.2 million people die on the road every year around the world. That's around 39 million people who've died on the roads around the world since I was born — equivalent to a city the size of Tokyo or New Delhi getting wiped off the map. The status quo is simply unacceptable, and something has to be done to fix it as soon as possible. Of the 1.2 million people that will die on the roads this year, about 40,000 will be Americans. That's about 3%. So if we moved entirely to self-driving cars in America and brought crashes down to 0, 97% of the world's crash fatalities would still be taking place as usual. Deploying a $200,000+ retrofitted self-driving car may work in a few American cities, but it is not going to make sense in most places around the world where fares are much cheaper. Most often, the choice is not between LIDAR and vision. It's between vision or nothing. The best system is the system that's there running on my car when I need it to save my life. To say that all self-driving cars must have LIDAR is to sentence most of the world to death. We can't write off computer vision if we want to make a serious dent in this problem. It's going to be a key piece of the solution. Let LIDAR based players build the best self-driving car they can, and let vision based players do the same. We need to be trying everythingshow more

Updated the subagents system in my custom Pi extension... to include: - Pi subagents - Codex subagents - Claude Code subagents The Pi thread can call and spin up any of them, and instead of just using the "codex exec" or whatever, it'll do a tool call that wraps it up in a first class way. Have a decent UI for them already added in as well. Pi being the main thread then having Codex/Claude Code as tools when needed feels so good. A flow I'm already using a ton is: "Implement feature ___, have a cc fable subagent research and plan out the api design + feature, then have a codex subagent implement it, then have a cc fable subagent do a review/fixes focused on simplicity and correctness" I also have a dynamic workflows extension setup and working, but it's not nearly as polished yet. Probably gonna make the subagent system more robust, then built it around that so u could have a workflow that flows between codex/cc/pi...show more

People still don't understand how good Codex has become... 👀 I asked it to build this Starbucks website and within some 30 minutes it generated this without me touching a single line of code or spending time looking for assets either. Imagen generated every visual and GPT hooked everything into the codebase. After a couple of small tweaks it all just worked. It didn't even use half of my Codex session. It's super surprising to see quality of the assets. Was way better than I expected. That made me think about what AI is actually changing. Being a developer or a designer was never just about doing the work, writing code or designing it using some kind of tool. It was about knowing what to build in the first place. AI is getting incredibly good at the doing. The hard part is still having good ideas, making the right decisions, and knowing when something feels right. That is only getting more valuable. The people who win won't be the ones writing the most code. They'll be the ones with the best ideas who know how to turn them into reality with AI. Live: Code:show more

BREAKING: Satellite images taken today (It is not posted... here) indicate that the IRGC Aerospace Force has successfully destroyed another component of a U.S. Army Patriot PAC-3 air defense system deployed at Erbil International Airport. The fire observed at the airport was caused by the destruction of one or more launchers belonging to the system at 36°15'01.52"N 43°57'52.94"E. A few days earlier, another Patriot PAC-3 launcher at the airport was also destroyed. It is believed that both attacks were carried out using Shahed-136 one-way attack drones launched by the IRGC Aerospace Force. One of these drones was also shot down over Erbil, Iraqi Kurdistan, earlier today. #OperationEpicFuryshow more

If you have a girlfriend who is into fashion,... or had one years ago, she probably had a Vinted account. Vinted was back in the day the most unhinged female-gaze-focused site on earth. It was 99% women talking about fashion and other stuff. I think today it is a little more professional. The site is basically like eBay for clothes. It was created by a girl who was annoyed by creepy private messages on eBay and wanted a safe space for females. Anyhow, Vinted was the anti-X. On X, it’s an all-men audience having a completely comical understanding of the female gaze, and on Vinted, you basically had young women thinking men are way more emotional than we are. People who form their opinions on women based on Reddit posts, Pornhub, and incels on X are just ridiculous copers. You can always tell by them being completely one-sided and sexualizing everything. A guy on X would be like, “Haha, doesn’t she know that men don’t care about fingernails? She is so dumb.” Bro, women do their fingernails like you wrap your car. It’s not meant to impress the other gender. It’s also ridiculous when incels threaten women with, “He will pump and dump you.” They don’t care. To them, it’s just an adventure. Sex gets increasingly more important to you the harder it is for you to get. Also women understand intersex dynamics way better. They get feedback from an early age. They get complimented, approached, harried based on what they wear. Incel men get no feedback at all until they have their first gf. That’s why their styling is always way off when they try to impress women.show more

Many of you might think face wash is just…face... wash. We use it to clean the dirt and grime from our daily lives. Some take it seriously while others just use it as a checklist of hygiene. But what if face wash was something more? What if there was a way to gently and quickly combat dry skin every day? Introducing H2.O Facial Cleanser by biöm – a groundbreaking product in the world of skincare. In my quest to redefine what a cleanser can do, I've created something extraordinary. Skin gets dirty, but it also gets thirsty. H2.O isn't just about cleaning your skin; it's about transforming it. The secret sauce in my unique formula is ElfaMoist, a hydration powerhouse. This active ingredient goes beyond the superficial layer of your skin, delivering hydration that penetrates deep into your pores, rivaling those satisfying Gatorade commercials. You can think of ElfaMoist like water that went to an Ivy League school. It’s clinically proven to hydrate your skin for over 30 hours, and it outperforms every other hydrator on the market. H2.O isn't just a cleanser; it's a hydration revolution. ⚠️ No fragrances. 🫒 Gentle olive & coconut oil-based surfactants. 🌿 Natural aloe vera for soothing hydration. 💧Squalane to mimic your skin’s natural oils. With each use, H2.O promises not only to cleanse but to leave your skin feeling refreshed, nourished, and more hydrated than ever before. It's time to experience a cleanser that does more than just clean – it rehydrates, replenishes, and renews. Welcome to the future of skincare.show more

CPN❗so apparently xzs blurred a ring throughout the whole... vlog 👀🤡 also he was wearing the same ring and the same outfit while he was on his way back to BJ after wrapping up filming for the ♾️ 10 days (so i guess he went straight to rehearsals 🥺) and judging by its thickness, it seems more like a fitness ring rather than a normal ring? ...cpn ! also have you noticed 🦁 is more of a watch guy ⌚ while 🐰 doesn't really wear watches much. He has always preferred rings 💍 than watches lol so our rock climber who goes around promoting the wh00p band couldn't convince 🐰 to wear one, so instead they went for something 🐰 is more comfortable with, which is a ring 💍... so a fitness ring? (so it makes sense why they would blur it instead of giving it a free promo) gonna file this away in caring boyfriends folder 📁🥹show more

What is Shiba Inu’s Shibarium? Shibarium is the Layer... 2 network built around the Shiba Inu ecosystem, designed to make transactions faster and cheaper than using Ethereum directly. Shibarium moves much of the activity away from Ethereum while still using Ethereum as its underlying settlement layer. SHIB itself remains an Ethereum-based token, separate from Shibarium, which is a distinct chain built to support the wider ecosystem. (1) It is designed to reduce transaction costs for Shiba Inu users and developers, making frequent onchain activity more practical. (2) It gives developers a dedicated network for building Shiba Inu ecosystem apps, including DeFi protocols, games and other Web3 applications. (3) ethereum:0x9813037ee2218799597d83d4a5b6f3b6778218d9 serves as Shibarium's gas token, its governance token through the Doggy DAO, and the asset validators and delegators stake to secure the network's proof-of-stake consensus. (4) Shibarium is also designed to expand the utility of the broader Shiba Inu ecosystem by moving activity beyond simply holding or trading $SHIB. (5) A portion of Shibarium's base transaction fees is converted into SHIB burns, an active mechanism running since block 6,206,570 that links network usage directly to SHIB's supply. Shibarium is the infrastructure layer intended to support a broader ecosystem of applications and onchain activity. As adoption grows, the key question will be whether Shibarium can turn its large community into sustained network usage.show more

A cat walked by my house last night and... immediately came in looking for food and water. It was thin, dirty but very friendly. My first thought, this has to be a house cat who lost its way. Then I saw it's ears, and hold up, this looks like a purebred cat. So I brought it to the vet to confirm, maybe it had a loving owner who put in a microchip, maybe it was already spayed/neutered. Unfortunately, the vet confirmed it's a female, less than 2 years old but already had babies. And it is an American curl. Most likely, a backyard breeder used her as a kitten mill, then threw her out when there was a problem? This is the heartbreaking reality of the irresponsible practices of backyard breeders. And why we have so many strays without homes. STOP SUPPORTING BACKYARD BREEDERS!!! ADOPT DON'T SHOP!!! By naddinkyshow more

Fine-tune DeepSeek-OCR on your own language! (100% local) DeepSeek-OCR... is a 3B-parameter vision model that achieves 97% precision while using 10× fewer vision tokens than text-based LLMs. It handles tables, papers, and handwriting without killing your GPU or budget. Why it matters: Most vision models treat documents as massive sequences of tokens, making long-context processing expensive and slow. DeepSeek-OCR uses context optical compression to convert 2D layouts into vision tokens, enabling efficient processing of complex documents. The best part? You can easily fine-tune it for your specific use case on a single GPU. I used Unsloth to run this experiment on Persian text and saw an 88.26% improvement in character error rate. ↳ Base model: 149% character error rate (CER) ↳ Fine-tuned model: 60% CER (57% more accurate) ↳ Training time: 60 steps on a single GPU Persian was just the test case. You can swap in your own dataset for any language, document type, or specific domain you're working with. I've shared the complete guide in the next tweet - all the code, notebooks, and environment setup ready to run with a single click. Everything is 100% open-source!show more

I still don't understand why everyone is not using... this yet. Thanks to it, a year ago I increased my income to 17,000 dollars a month Andrey Karpathy, co-founder of OpenAI, published a simple idea that got 16 million views: stop using AI to write code, use it to build a second brain You point Claude Code to a folder, drop any source in there (an article, transcript, PDF) and Claude reads it, links it, and saves it into a living wiki of everything you know. It compounds like interest: the more you feed it, the smarter it gets Here is the gist: Install Obsidian, create a vault, open it in Claude Code Paste the file with Karpathy's wiki idea and tell Claude to build it Claude creates three folders: raw for sources, wiki for its pages, CLAUDE which runs everything Drop any source into raw and say "ingest this" Ask questions across everything, forever Five minutes to set up, and you never start with an empty chat again The full step-by-step guide is in the article. Save to bookmarksshow more

Something important to mention is that zoink is being... extremely accusatory towards me regarding my playtime and attempts, and many people are going to side with what he is saying simply because of his platform. He feels the need to spread misinformation about me to get even more praise than he already does. Regarding my flamewall and heliopolis playtime, i have made it very public that i had old playtime on flamewall and it has also always been public information. The tatox3 video coaching thing is something that i addressed in the completion description as well as just being public information i never made an attempt to hide. Popoff was also very aware of my previous playtime and still agreed to race because it was purely for fun and motivation. As for heliopolis, i started playing this level the night before i started streaming it to decide whether or not i actually wanted to go through with it. i made this playtime completely public at the beginning of my stream and did not lie about anything. i attached a clip for you guys to see that from the beginning of playing the level i made all of my previous playtime extremely clear. Zoink is trying to twist my wording of "day 1, 2 3," etc to make it appear as if ive been lying but my playtime has always been clear.show more

Did you know? This mission was born in a... bar and it was sketched on a cocktail napkin Two scientists and a Lockheed engineer sat in a Tucson bar, not far from University of Arizona campus. Their idea was rejected by NASA, TWICE for being impractical and too expensive. The spacecraft touched the asteroid surface for only "5 seconds" and collected more than enough samples using a device called, TAGSAM (Touch-and-Go Sample Acquisition Mechanism) which is an engineering marvel. But how does it work? When it contacts the asteroid surface, it releases high-pressure nitrogen gas to fluidize loose material, and capture ✨stardust ✨. Read more about it in my bog, linked below.show more

After a few more hours, I think I've figured... out Opus 5. Opus 5 is trained to be more agentic than anything I've used. All Claude 5 models are like that. So what changes? The way to interact with Opus 5 or contextualize it won't work the same way as with other models. It loves exploring, so it doesn't need much guidance for it. Unique preferences, artifacts, and references compliment it well and enable cleaner and more effective exploration and execution. Now that it can explore more effectively on its own and understand intent better, the best thing to do is to get out of its way (e.g., it doesn't need examples of your preferences; a clear high-level description of it works best). It's truly agentic in that sense. A good first step to provide better context for Opus 5 is to distinguish between what's situational and what needs persistence. Regardless, persistent system prompts and CLAUDE.MD needs to stay lightweight. Remove memories and tool descriptions from these. CLAUDE.MD is also a great place to tap into progressive disclosure by linking command/skills to it. On the situational side, agent skills and auto-memory can leverage progressive disclosure and the improved ability of the model to use its external context/knowledge. Conflicting and unnecessary instructions, which are common at this layer (mainly to ensure reliability), are going to throw off this model easily. That's the biggest change I had to make. Simple, clean, and clear prompts and skills work best. I had to clean a lot of my skills and system prompts. The way I prompt remains the same (usually clear and well-scoped). MCP tool descriptions are also more descriptive and have been deduped from the system prompt. Anthropic released a guide on the new rules for context engineering, which was helpful here. I started to test the recommendations and created a little artifact with the things that worked along the way. This might feel like a lot of work. Believe me, it has been frustrating. But I think we can expect future frontier models to become more agentic and smarter at figuring out the right context/gaps. The best thing to do is to prepare for that now. Boris Cherny mentioned that Opus 5 is their least prompt-injectable model yet. I am not sure if that was something they intentionally trained for or if it emerged based on how it was trained, which is to be extremely agentic in nature and more direct in execution.show more