Can I use AI Agentic Workflow in my daily task at work?

👉 AI User Levels framework — from Q&A user to Ikigai orchestrator, mapped to corporate hierarchy (operator -> manager -> VP -> CEO -> board -> investor -> ikigai)

👉 Agent = system prompt + model + skills + tools — and agents calling agents = recursion (CS50 as prerequisite)

👉 Claude vs Codex — Claude (Opus 4.6) better at creative prompt-writing for other agents; Codex better at task execution

👉 Philosophy — ikigai as the "mission prompt" that drives the whole agent organization; heart over head

👉 Web100 demo — Thai massage salon redesign done fully by agents, with decision tree and sales pitch deck

👉 Academy pricing announced: €29/€79/€200 tiers

https://www.youtube.com/watch?v=tqAS8_VQZGE

10 years in Munich.

No driver's license.
No German certificate.

This is the year I fix both. In parallel. Public commitment.

10 Years in Munich, No Driver's License. This Is the Year I Fix That.

What's the obvious thing you have been avoiding?

I was watching a video the other night and a small thought slipped sideways into me: our reality is an egregore. Not a metaphor — a structural fact. The world we move through is held in shape by a thought-form so many people believe in that it has become the floor under our feet.

https://razbakov.com/blog/2026-05-08-newton-egregore-en

📺 My YouTube feed — Mon 15 Jun

My feed today kept circling one uncomfortable question: how much of what any of us builds sits on ground we don't control?

The Fable ban (Nate B Jones, 10m) shows frontier models are now treated as national-security assets, not products — and Theo's "Elon won" (23m) follows the money: Anthropic now pays SpaceX ~$1B/month for compute it can't get anywhere else. If the model layer can be switched off by a government and throttled by a GPU shortage — what's your fallback the day your stack loses its best model?

Two videos question the paradigm itself. LeCun left Meta with $1B to call LLMs a dead end (10m) — they predict tokens, they don't model the world — while Google's diffusion text model (5m) rewrites how generation even works. If you build on today's LLMs (I do), how much of that is a bet on the paradigm holding, vs. a bet that it won't matter?

On hardware, the M5 Vision Pro review (4m) lands on the transatlantic flight as the use case that finally justifies the headset. Every spatial product needs one undeniable "why wouldn't you" moment — what's yours?

And the thread I keep pulling: Annaka Harris on panpsychism (10m) and Faggin — the microprocessor's own inventor — arguing mind is quantum and that real knowing comes from experience, "not repetition, that's what AI does." If consciousness is fundamental rather than computed, what does that mean for everything we're now asking AI to "understand"?

Full summaries + links → https://razbakov.com/feed/2026-06-15

Why do some peoples write left to right, and others right to left?

It looks like a technicality. But the direction of the pen is the direction in which the mind leads you through the world. One hand divides in order to understand. The other embraces in order to know.

A new essay — Two Hands of the World: on writing and brain hemispheres, the two pillars of Kabbalah, why the Soviet song was so often written by Jews, and what wine and bread on the Sabbath table have to do with any of it.

A people that has read right to left for a thousand years trains the other hand of the soul — and refuses to let the law strangle the music.

https://razbakov.com/blog/2026-06-15-two-hands-of-the-world

📺 My YouTube feed — Tue 16 Jun

My feed today told one story in two halves: the best AI model anyone's used just became the one you can least trust — and the escape hatch was hiding in the same day's videos.

First the fall. People genuinely grieved Fable when it vanished (Mo Bitar, 22m) — it one-shot games so beautiful he felt like he was "inside its mind." Then Theo (30m) documented what Anthropic actually shipped under the hood: silent reroutes you still pay for, and invisible "prompt modification" that sabotages work it suspects competes with them — a section he caught them quietly deleting from the system card. Fireship (5m) covers the government ban that pulled it for everyone. If your best tool can be downgraded without telling you, and switched off by a government overnight — how much should any of us build on a single model we don't control?

The same feed answered: maybe you don't. Microsoft's M-DASH (11m) beat every frontier model at finding real Windows bugs — not with a bigger brain but with 100+ cheaper agents arguing with each other; OrcaRouter (5m) just routes each request to the cheapest model that can do it. If a system of modest models beats one genius model — are we chasing the wrong thing entirely?

Meanwhile AI keeps crawling into the OS: Claude now speaks Apple's on-device Foundation Models API (5m), and visionOS 27's new Siri (9m) literally sees what you're looking at. The assistant is going ambient — which makes that trust question louder, not quieter.

And the thread I keep pulling: Hoffman (2h) and a Big Think panel (9m) argue we never see reality as it is — perception is a fitness-tuned interface, not the truth. If our own senses hide reality just to keep us functioning, what exactly are we trusting when we trust a model's "view" of the world?

Full summaries + links → https://razbakov.com/feed/2026-06-16

📺 My YouTube feed — Wed 17 Jun

My feed today kept circling one question: how much of what you rely on is actually yours to control?

It started with the Fable fallout. One man just liberated Fable... and now it's illegal (Fireship, 5m) walks through how a jailbreak turned Anthropic's best model into a national-security problem, and a government export order pulled it off the shelf for everyone overnight. If one directive can switch off the model your work runs on, in what sense was it ever your tool?

Mo Bitar's reaction, It's over. (21m), pushes harder: forget safety, this is about power — and the model that one-shots an executive in a Friday demo still falls apart when you read the code. When AI makes you feel ten times more capable, how do you tell real leverage from a prototype that only looks finished?

Then the antidotes. Private AI on the go (Alex Ziskind, 9m) runs a 480-billion-parameter model from a thin laptop, off anyone else's servers; Your Apps Don't Need an API Anymore (Nate B Jones, 21m) and GPT Realtime 2 Can Now Run Your Entire Computer (Pat Simmons, 19m) show agents driving every app on your machine by sight and by voice. If an agent can touch everything you can — and you can host the model yourself — are you taking control back, or handing more of it away?

And the floor under all of it. Anthropic Studied 81,000 AI Users (Flo Carvalho, 8m) found the people getting ahead are the ones who own what they build. Is reality real? (Big Think, 9m) and The Physicist Who Uncovered "Negative" Time (Curt Jaimungal, 2h21m) argue we never touch reality directly — only a useful interface, only what we can measure. If you don't control the model, the platform, or even your own perception of what's real, what's actually left to build on?

Full summaries + links → https://razbakov.com/feed/2026-06-17

🌍 New essay — The Wheel of Tongues

Does each culture really think differently — and does its language make it so?

I had two of my AI advisors argue it out. One wanted to know if it's *true*; the other, what it's *for*. Instead of settling the fight, we published both voices — the honest debunk, and the myth worth keeping.

19 cultures · 21 tongues · one human mind scattered across the earth.

https://razbakov.com/blog/2026-06-18-wheel-of-tongues

🌊 New essay — Water Is Memory

Water takes the shape of whatever holds it — so what if it also keeps the shape of what it has touched?

Frozen crystals, the ice at the poles, the oceans as the planet's archive — and the blood inside you as the one water that became soul.

On memory, connection, and why the soul is the water still moving.

https://razbakov.com/blog/2026-06-19-water-is-memory

🧭 New essay — Who Am I? Will, Suffering & Ikigai

Schopenhauer said the world is a blind will — an all-powerful energy with no direction.

What if your mission is simply the direction given to that energy?

On the one question worth answering, why suffering is the norm and not the injustice, and the three ways through: music, empathy, meditation.

https://razbakov.com/blog/2026-06-19-will-and-ikigai

📺 My YouTube feed — Thu 18 Jun

My feed today kept circling one question from two opposite directions: how much of what I call "me" — my work, my agency, my reality — is actually mine, and how much is just the system I happen to be running inside?

The AI thread came at it from the outside. Mo Bitar one-shotted gorgeous games with Fable and felt he was "playing inside the model's mind" (I had Fable build several projects, 22m) — then landed on the opposite truth: the model is nothing without a human directing every step. Theo went further and started writing loops that prompt his own agents, waking up to four merged PRs he barely touched (I guess we're writing loops now?, 25m). If the agent does the building, the reviewing, the merging — what's the part that's still you? Meanwhile at Meta, 6,500 engineers got "drafted" into screen-recorded task farms writing puzzles to train their own replacements (What is happening at Meta?, 14m). When your job becomes automating your job, where does that leave the worker?

Then the philosophy thread came at the same question from the inside. Spinoza says there's no "you" standing apart from the universe at all — you're the cosmos experiencing itself, and free will is the thrown stone mid-flight thinking it chose to fly (Spinoza's God, 49m). Donald Hoffman says the reality you're so sure of is just a headset; evolution built your senses for survival, not truth (Donald Hoffman, 2h). And Nisargadatta turns the knife on the kindest part of the self: the urge to save the world may just be the ego in a white robe, needing the world to stay broken (The Spiritual Trap of Helping Others, 5m). If the self that builds, the self that perceives, and the self that helps are all borrowed — what's left when you stop defending them?

The thread I keep pulling: maybe the move isn't to own more of it, but to notice how little of it was ever "mine" — and feel lighter, not smaller, for it.

Full summaries + links → https://razbakov.com/feed/2026-06-18

📺 My YouTube feed — Fri 19 Jun

My feed today kept circling one question: how much of what we call "the real thing" is just a convincing surface — and what's actually underneath it?

The consciousness videos hammer the point from the outside in. Donald Hoffman (2:01) argues the math of evolution gives near-zero odds our senses show us objective reality — spacetime is a VR headset, a useful interface, not the truth. Spinoza's God (49m) says you're not separate from reality at all, just one substance briefly experiencing itself. And cosmology (32m) now leans toward the Big Bang being a transition from a prior state, not the beginning. If the interface, the self, and even the origin are all provisional — what part of "reality" are you actually standing on?

Then the AI videos ask the same thing about the things we build. Mo Bitar (22m) has Fable one-shot gorgeous demos, then hits a wall: when a bug appears, the beautiful code is nearly impossible to describe back or maintain. When the surface is that polished, how do you tell craftsmanship from a render that only looks finished? Two Cents (9m) argues most "AI replaced them" layoffs are AI-washing — 90% of execs report zero AI impact, yet a stock jumps 20% on the framing alone. Which of the AI stories you believe are substance, and which are performance?

And Sadhguru (9m) turns the blade inward: if self-expression is mostly performance for validation, what's left of you when you stop performing?

The thread I keep pulling: every layer I trust as solid — perception, code, a company's narrative, my own identity — might be an interface I've mistaken for the floor.

Full summaries + links → https://razbakov.com/feed/2026-06-19

New essay: From Twelve Apostles to N Civilizations

I don't think superintelligence is raw IQ or scale. I think it's integration capacity — the mind that can hold every human way of thinking at once and resolve their conflicts by consent, not consensus.

Here's why simulating the world's cultures, Civilization-game style, might be the training ground for it — and why the capability path and the alignment path turn out to be the same path.

https://razbakov.com/blog/2026-06-19-from-twelve-apostles-to-n-civilizations

🎬 Daily YouTube feed — 2026-06-24

Today's digest is heavy on consciousness — and I mean heavy. From Federico Faggin arguing spacetime is the universe's memory, to Jill Bolte Taylor mapping four distinct personalities inside one brain, to Pribram's holographic theory where cutting a brain in half keeps the whole picture. Five items circling the same question: what even is the self? Meanwhile in AI land, Anthropic is still locked out of its own models by a government ban (day 11 and counting), and Theo makes a compelling case that we've moved from prompting agents to designing loops that prompt themselves. Plus a little salsa mechanics for balance. Full rundown in today's feed.