Hkn

For those new to AI or who think ChatGPT is the only chatbot, here are the top proprietary and open-source AI chatbots. Try them all and you'll be amazed at the intelligence at your fingertips. Proprietary AI OpenAI (ChatGPT): chatgpt.com Anthropic (Claude): claude.ai Google (Gemini): gemini.google.com xAI (Grok): grok.com Perplexity AI: perplexity.ai Open-Source AI DeepSeek: chat.deepseek.com Moonshot AI (Kimi): kimi.com Meta (Llama): meta.ai Mistral AI: chat.mistral.ai Alibaba (Qwen): chat.qwen.ai But if you have to use just one, use Anthropic Claude claude.ai Bookmark this and check them out.

We are making our discount permanent! 🎉 Enjoy building with DeepSeek-V4-Pro and bring your innovative ideas to life! 🚀

@AbdullahIs74903 And may Almighty Allah forgive Yunusa and bless him with al-jannah.

My first wife married me when I had absolutely no source of income. We decided to marry early to avoid zina due to my high libido. I was honest with her from the beginning about my financial situation, but she insisted we get married anyway. After the Nikkah, she moved into my parents' house in Oyo State, while I relocated to Ogijo, Ogun State, to hustle. I saved some money for a year from where I was working at the MRI Metal Recycling Industry located in Ogijo, Ogun State. During my hustling time, my wife came to where I was renting to spend a week with me. From there, she got pregnant for me. Afterward, I moved back to Oyo State completely and started fishery business. The first fish I reared, I couldn’t see my capital, let alone gain. Depression filled me, but my wife always gave me joy with her comforting words. A friend of mine who passed away—his name was Yunusa—was the one who dashed me N1million to start my business again. Today, I am the owner of plenty of ponds with workers. May Allah bless all the good wives who married their husbands when they had nothing.🙏🏼🙏🏼

@0xahmard @AbdullahIs74903 Ameen

@AbdullahIs74903 May Allah forgive Yunusa shortcomings and make his grave more comforting than his bed in this life

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.

// Harnessing Agentic Evolution // Pay attention to this one if you run iterative agentic search loops. (bookmark it) AEvo splits the self-improvement loop into two jobs: > One proposes the next candidate. > The other watches what worked, what failed, and edits the procedure that proposes future candidates. Past runs (candidates, feedback, traces, failures) become memory the meta-agent reads from. Achieves 26% relative gain over the strongest evolution baseline on agentic and reasoning benchmarks. SOTA on three open-ended optimization tasks under the same iteration budget. If you are accumulating agentic search logs you never use, this is how to feed them back into the search procedure itself. Paper: arxiv.org/abs/2605.13821 Learn to build effective AI agents in our academy: academy.dair.ai

@abdool_moh Safe trip bro. You are doing a great job 👍

Out to seal another car deal, may the almighty see us through.

In the next 500 years, it’s either you are enjoying or suffering in the grave waiting for judgment day. The choice is yours now.

Really cool work from the team reimagining the mouse pointer to be intelligent! Try the prototype in @GoogleAIStudio it's pretty magical.

Godfather of AI: "If you sleep well tonight, you may not have understood this lecture." This 47-minute lecture is the best thing I saw about AI in the last few months. It will definitely help you understand how it actually works and where it's going. Geoffrey Hinton built the neural networks behind every AI alive, then quit Google to warn the world about it. The part nobody wanted to hear: > AI is already developing abilities its creators didn't intend > in most cognitive tasks it's already ahead of us > the question is no longer if it surpasses us but when > the only decision left is which side of that line you're on Right now the average person opens Claude, types something, gets an answer, closes the tab. They think they're using AI. they're using maybe 10% of it. I went through his entire lecture, built a practical system from what he was describing. 18 steps to actually use Claude the right way, with copy-paste prompts that work today. Full guide in the post below.

Demis Hassabis says he can cure every disease in 10 years. Most people roll their eyes when they hear this, but I don't. Demis is the guy who just won the Nobel Prize for solving protein folding with AI (a problem biologists had been stuck on for 50 years). But that was just one milestone in his much grander plan. In 2010, he founded DeepMind with a 2-part mission: "solve intelligence, then use it to solve everything else." Step 1: make AI good enough to do real science. Step 2: point that AI at humanity's biggest problems. Step one was AlphaFold. He used AI to figure out the 3D shape of every protein in nature (which is basically what every drug attaches to). Demis said it would have taken "a billion years of PhD time" to do by hand. Step two is curing all disease. And as of today, step two is fully funded. Isomorphic Labs (his AI drug discovery company inside Google) just raised $2.1B led by Thrive Capital. Here's where the money goes and what Demis thinks happens next: > Drug discovery currently takes 5-10 years and costs billions per drug. That math is why most diseases don't have good treatments today. > AI fixes the math. Their drug design engine compresses development from years to months. Maybe weeks. > Isomorphic's first AI-designed cancer drug enters human trials this year. > Their pipeline expands beyond the current 17 programs across cancer, immune diseases, and heart disease into more health domains. > The endgame is personalized medicine: drugs designed overnight for your specific biology and your specific disease. That last one is the whole point. Today's drugs are mass-produced for an "average" patient who doesn't really exist. So most existing treatments work inconsistently from person to person, and most rare diseases never get a treatment at all (no market = no drug). When drug design gets fast and cheap, that whole calculus flips. Cancer variants get drugs designed for that specific variant, rare diseases get treatments because economics stop mattering, and drug-resistant infections get new drugs faster than they can evolve. That's what curing every disease actually looks like. Now imagine what your life looks like in 2036. A doctor draws your blood, sequences your genome, sends your disease profile to an AI. By morning the AI has designed a custom drug for your specific biology. Side effects, dosage, drug interactions all worked out before you take the first pill. You and your kids never see a cancer ward. That's what $2.1B is buying today. Demis was right about AlphaFold. If you consider the possibility that he's right again, every disease alive today is on borrowed time.

“ELF: Embedded Language Flows” Most diffusion LMs still treat language as discrete tokens, or keep forcing continuous states back into token space during generation. This paper runs diffusion almost entirely in continuous embedding space, then convert to tokens only at the final step. This makes language diffusion look much more like image diffusion, so tricks like Flow Matching, stochastic sampling, and classifier-free guidance become natural. ELF beats strong discrete and continuous diffusion LMs with fewer sampling steps and about 10x fewer training tokens.

The future of Local LLM is so bright.

Reinforcing Recursive Language Models Can a 4B model learn to recursively call itself to answer hard long-context questions? We RL fine-tuned a small model to behave as a native RLM. On evidence selection across scientific papers, our 4B RLM matches Sonnet 4.6 in quality while running significantly faster and cheaper.

Today we release Token Superposition Training (TST), a modification to the standard LLM pretraining loop that produces a 2-3× wall-clock speedup at matched FLOPs without changing the model architecture, optimizer, tokenizer, or training data. During the first third of training, the model reads and predicts contiguous bags of tokens, averaging their embeddings on the input side and predicting the next bag with a modified cross-entropy on the output side. For the remainder of the run, it trains normally on next-token prediction. The inference-time model is identical to one produced by conventional pretraining. Validated at 270M, 600M, and 3B dense scales, and at 10B-A1B MoE. The work on TST was led by @bloc97_, @gigant_theo, and @theemozilla.

People talk, listen, watch, think, and collaborate at the same time, in real time. We've designed an AI that works with people the same way. We share our approach, early results, and a quick look at our model in action. thinkingmachines.ai/blog/interacti…

"Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior" Most activation steering treats concepts like directions, but model representations are often curved, so straight line steering can cut through unnatural internal states. The key idea of the paper is to steer along the activation manifold instead of through it. Across LLM reasoning tasks and a visual world model, manifold steering produced smoother and more natural behavior, while linear steering caused probability mass or visual states to jump abruptly.

// LLMs Improving LLMs // Interesting progress the past of couple of weeks around self-improving AI agents. If autoresearch was interesting, you will like this read. (bookmark it) We've been hand-tuning test-time scaling for a year. This work asks what happens when you let an LLM search the space instead. The paper introduces AutoTTS, a framework that reframes the human role: instead of designing branching, pruning, and stopping heuristics directly, you construct a discovery environment where TTS strategies can be searched automatically. They formulate width–depth TTS as controller synthesis over pre-collected reasoning trajectories and probe signals, so candidate controllers can be evaluated cheaply without repeated LLM calls. Two design choices carry the search. Beta parameterization makes the control space tractable. Fine-grained execution-trace feedback tells the explorer LLM why a candidate failed, not just that it did. On math reasoning benchmarks, the discovered controllers beat strong hand-designed baselines on the accuracy–cost Pareto frontier and generalize zero-shot to held-out benchmarks and model scales. Entire discovery cost: $39.9 and 160 minutes. Why it matters: The era of researchers hand-crafting CoT, best-of-N, and self-consistency recipes is on a clock. Once the search loop is cheap enough, TTS becomes another thing LLMs do for themselves. Paper: arxiv.org/abs/2605.08083 Learn to build effective AI agents in our academy: academy.dair.ai

New in Claude Code: agent view. One list of all your sessions, available today as a research preview.

This works really well btw, at the end of your query ask your LLM to "structure your response as HTML", then view the generated file in your browser. I've also had some success asking the LLM to present its output as slideshows, etc. More generally, imo audio is the human-preferred input to AIs but vision (images/animations/video) is the preferred output from them. Around a ~third of our brains are a massively parallel processor dedicated to vision, it is the 10-lane superhighway of information into brain. As AI improves, I think we'll see a progression that takes advantage: 1) raw text (hard/effortful to read) 2) markdown (bold, italic, headings, tables, a bit easier on the eyes) <-- current default 3) HTML (still procedural with underlying code, but a lot more flexibility on the graphics, layout, even interactivity) <-- early but forming new good default ...4,5,6,... n) interactive neural videos/simulations Imo the extrapolation (though the technology doesn't exist just yet) ends in some kind of interactive videos generated directly by a diffusion neural net. Many open questions as to how exact/procedural "Software 1.0" artifacts (e.g. interactive simulations) may be woven together with neural artifacts (diffusion grids), but generally something in the direction of the recently viral x.com/zan2434/status… There are also improvements necessary and pending at the input. Audio nor text nor video alone are not enough, e.g. I feel a need to point/gesture to things on the screen, similar to all the things you would do with a person physically next to you and your computer screen. TLDR The input/output mind meld between humans and AIs is ongoing and there is a lot of work to do and significant progress to be made, way before jumping all the way into neuralink-esque BCIs and all that. For what's worth exploring at the current stage, hot tip try ask for HTML.

KIMI FOUNDER JUST DROPPED A 40-MINUTE MASTERCLASS. The exact architecture behind a $20B valuation — there's no faster way to learn how to build AI agents right now. Bookmark this for the weekend. 40 minutes. zero fluff. from the person who built it. Optimization → Linear Attention → Sub-Agents → Open Systems → Cash