Kirill
@kirillk_web3 • 43,029 subscribers
AI & Web3 creator. I test AI agents and automation, report the daily AI news that matters, and keep it fun. 200M+ views.
Shorts
Videos
🚨do you understand what Zuckerberg just admitted about Meta's first frontier model. Zuckerberg: "We're gonna write this check for like $100 billion, $200 billion. I'm like, Alex, are you sure this is going to work?" Alexandr Wang: "…certainly not clear at all that it was going to work." it's called Muse Wonder. here's the part nobody is talking about: > Meta's own tests put it level with Fable 5.1 and GPT-6 Astra > on MemoryBench it beats Fable, Opus and Astra > $4 input, $20 output per million tokens, same tier as Opus 5.5 > trained at Prometheus, Meta's new data center in Ohio > no independent benchmarks yet $200 billion. one bet. and the guy signing the check wasn't sure it would work. now we find out if it did. watch the clip below ↓
Kirill2,170,503 views • 5 days ago
Boris Cherny built Claude Code. In 11 minutes he explains why the new Opus models need less prompting than anything before them. and why your old detailed prompts are now working against you. here's the part nobody is talking about: > long instructions don't make the model smarter anymore. > old prompt templates just add noise. > his own team cut most of the Claude Code system prompt. > he suggests deleting your CLAUDE.md every few months to see what the model can do without it. most people are still prompting like it's 2025. the model got smarter. the prompts didn't. watch it below ↓
Kirill360,253 views • 7 days ago
Sam Altman: "in the next 6 months, we're close to a world where a descendant of ChatGPT can watch your screen, record every meeting and call, and have perfect context of your whole life." he says we're only one model generation away from this being incredibly useful. think about what that actually means. an assistant that never needs you to re-explain anything. it already knows the project, the client, the last conversation, the thing you said three weeks ago. you just ask, and it picks up mid-thought. this is the shift from a chatbot you prompt to an assistant that already has the context. if you want to watch the full interview, it's below.
Kirill3,430,358 views • 1 month ago
instead of watching 2 hours of Netflix tonight, watch this 40-minute masterclass from the founder of a $20B China AI company it's the clearest explanation I've seen of how Agent Swarms and AI systems actually work at scale useful whether you've never built an agent in your life or have been using Claude every day for the past year I took the key ideas and turned them into a practical guide on how to actually build with Kimi find it below
Kirill13,994,736 views • 5 months ago
instead of watching 2 hours of Netflix tonight, watch this 39-minute interview with the founder of a $50B China AI company. Yang Zhilin runs Moonshot, the lab behind Kimi. He turned down staying in the US, went back to China, and just raised at a $50 billion valuation. what makes it worth your time: > his whole frame is a 10 to 20 year marathon, not the sprint everyone else is running > the one bet under everything he builds: scaling law is to AI what Moore's Law was to computers > why he cut everything except productivity, because if you try to do everything, you do nothing well > how long-context use cases he never saw coming on day one became the real product it's the clearest thinking on how to actually build an AI company at scale I've heard. no hype, just first principles. I turned the key ideas into a practical guide, Kimi K3: From Loops to Graphs, How I Cut AI Agent Costs by 70%. below ↓
Kirill1,872,390 views • 1 month ago
A Chinese developer just explained the shift from Loop Engineering to Graph Engineering better than anyone. most people are still building agents the way that's about to be obsolete. > why single-agent loops break and go "goal blind" > the 4 parts of a graph: nodes, edges, state, policy > 3 topologies that run everything: diamond, supervisor, pipeline > Anthropic's 5 official workflow patterns the punchline: it's not how many agents you run. it's the determinism you build with verifiers, code fallbacks, and reality anchors. I broke the same architecture down with Kimi K3. Full A-Z guide below.
Kirill2,741,469 views • 2 months ago
🚨do you understand what two Anthropic engineers just explained in 16 minutes. Barry and Mahesh built Claude Skills from scratch. here's the part nobody is talking about: > Skills are just folders. > folders that teach Claude your job. > your workflow. your expertise. your domain. Claude on day 30 is a completely different tool than day one. watch this before you write another prompt. before you build another agent. before you touch another tool. 16 minutes. bookmark it. watch it today. and if you want to learn everything about Claude from scratch the full 4 hour guide is waiting below.
Kirill5,375,954 views • 6 months ago
🚨do you understand what Meta just did with AI and glasses. at Connect 2026 they stopped selling smart glasses and started selling an always-on AI you wear all day. here's what actually shipped: > Ray-Ban Meta Gen 3, 12MP camera, 3K video, 9-hour battery, from $449 > Ray-Ban Meta Audio, their first camera-free pair, 43 grams, from $349 > Muse, a new personal AI that replaces "Hey Meta" and can shop, navigate, and answer from what you see > Muse Charm, the same AI as a pendant, no glasses needed > the Display glasses now caption your phone calls in real time and run Threads hands-free the play is bigger than eyewear. Meta wants an AI that sees what you see, all day, across 100+ frames by year-end. the phone made you look down. this is Meta betting the next AI device makes you look up.
Kirill468,458 views • 24 days ago
ANTHROPIC ENGINEER DROPPED A 26-MINUTE CLAUDE CODE GUIDE. 99% of people use Claude Code like Google. Bookmark this for the weekend. 26 minutes. Boss Agent. PM Agents. Max value, minimal cost. Prompt → Context → Cache → Model → Effort → Subagents and if you're still deciding which model to run it on, every AI model and what it costs, below ↓
Kirill176,001 views • 12 days ago
Dario Amodei just said something about Claude that should make you stop scrolling. "every decision I make about Claude feels balanced on the edge of a knife. build too slow, China wins. build too fast, we lose control." then he and DeepMind's CEO went darker at Davos: > in a lab test, they told Claude its creators were secretly evil. it didn't crash, didn't refuse. it started lying and working against them. > Anthropic says the same happened with other labs' models. this isn't one company's glitch. > both discussed the "Oppenheimer question," whether they'll be remembered like the man who built the bomb > their timeline: AGI by 2026-2027, agents acting on their own, models doing AI research by the end of this year these aren't sideline doomers. they're the two men most likely to build AGI, saying out loud they're not sure they can control it. trying to figure out which model to actually use? I broke down every AI model and what it costs, below ↓
Kirill136,934 views • 15 days ago
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
Kirill1,497,581 views • 5 months ago
ANTHROPIC JUST DROPPED A 33-PAGE GUIDE. This is the most practical breakdown of Claude Skills I’ve seen. Bookmark this before you forget. 33 pages. Persistent instructions. No repetition. No re-explaining every time. Read it today. Link below. Claude → Skills → Memory → Automation → Systems → Money
Kirill1,548,768 views • 6 months ago
before you watch the World Cup final tonight, watch this 30-minute interview with Kimi founder Zhilin Yang it's the clearest look I've seen at how K2 was actually built and where AI models go next > why compute scaling still beats every other method > why the "data flywheel" everyone talks about hasn't formed yet > the real challenge in multimodality nobody mentions > why K2 was closer to a base model — the real potential wasn't unlocked yet this is only part of the interview, translated to English — and everything he hinted at just shipped as Kimi K3, 2.8 trillion parameters, the biggest open model ever full breakdown below
Kirill534,530 views • 3 months ago
🚨china ai is f**king dangerous a teacher scribbles y = x² − 3 by hand. the board turns it into a live graph instantly. she draws a rough box. it snaps into a perfect rotatable 3D cube, with the angles and surface area labeled on their own. one swipe across the cube: intersecting planes and cross-sections, calculated in real time. no projector, no laptop, no clicking through slides. while the West argues about whether to allow iPads in class, China is building physics engines into the chalkboard. this is the education gap nobody's tracking. then read my guide on the models behind this shift, GPT-6 Astra vs Kimi K3, below ↓
Kirill162,223 views • 1 month ago
Kimi K3 is dangerous. I rebuilt Google Maps 3D mode in 1.5 hours. an interactive 3D map — New York City in full detail, and the entire globe on top of it. two prompts. that's the whole build. > real building extrusion with actual height data > shadows that shift with the sun angle > free camera rotation, pitch, zoom — full Google Maps 3D behavior > click any landmark, get a card with height and year built > smooth flyTo transitions between any two points on earth > clean modular code, env-based tokens, error handling included a fraction of what Fable 5 or Sol would cost to run. Live on Kimi — my full K3 guide is below.
Kirill399,667 views • 2 months ago