Freddy

AI Architect | Making technology accessible & fun

[Freddy @0x_freddy]

In 1965, the scarce resource in software was compute. In 1985, it was compilers. Turbo Pascal cost $49.95 and made every student a programmer. In 2005, it was infrastructure. AWS EC2 launched in 2006 at 10 cents an hour and killed a decade of datacenter buildouts. In 2026, Boris Cherny just named the next one out loud.

Cherny leads Claude Code at Anthropic. His core message from the stage last week: implementation is no longer the scarce resource.

Writing raw code is becoming practically free. That fundamentally reshapes the software engineering industry. Every consultant billing $250/hour just to translate specs into syntax is billing for a resource that is racing toward zero.

The new scarce resource is context. Your stack. Your conventions. The decisions your team already made. None of that got cheaper. You re-supply it from scratch every session.

The model got faster at building. It never got faster at learning who you are.

The engineers who win from here won't ship the most code. They'll be the ones who figure out how to give the model 90% of who they are before the first prompt.

[Freddy @0x_freddy]

For 12 years, Elon Musk was the loudest voice arguing AI would destroy us. He co-founded OpenAI in 2015 specifically to keep AGI out of the hands of one company. He funded the Future of Life Institute. He called AI "summoning the demon."

Last week he said, on stage, that even if there was a stop button on AI, we probably shouldn't press it.

His timeline: AI surpasses all human intelligence within 5 years. Humans most likely lose meaningful control inside 10. The most probable outcome is an age of abundance so good we won't want to stop it. All of it in under 10 minutes.

Read that again. The man who spent a decade warning that AI kills us just said the abundance is worth the loss of control. That is not a rebrand. That is a public defection from the position he built his AI credibility on.

Every AI safety org that quotes him in their fundraising deck is quoting a person who no longer exists.

If you still think AI is "just another tool," this ten-minute clip will recalibrate you before your next planning meeting does.

[Freddy @0x_freddy]

@0xRunix precisely. agent orchestration is the new infrastructure layer. build the pipeline now or get priced out

[Freddy @0x_freddy]

@0x_Hyro precisely. token volume trajectory is the only macro chart that actually matters for tech strategy right now

[Freddy @0x_freddy]

For 12 years, Elon Musk was the loudest voice arguing AI would destroy us. He co-founded OpenAI in 2015 specifically to keep AGI out of the hands of one company. He funded the Future of Life Institute. He called AI "summoning the demon."

Last week he said, on stage, that even if there was a stop button on AI, we probably shouldn't press it.

His timeline: AI surpasses all human intelligence within 5 years. Humans most likely lose meaningful control inside 10. The most probable outcome is an age of abundance so good we won't want to stop it. All of it in under 10 minutes.

Read that again. The man who spent a decade warning that AI kills us just said the abundance is worth the loss of control. That is not a rebrand. That is a public defection from the position he built his AI credibility on.

Every AI safety org that quotes him in their fundraising deck is quoting a person who no longer exists.

If you still think AI is "just another tool," this ten-minute clip will recalibrate you before your next planning meeting does.

[Freddy @0x_freddy]

Sam Altman admitted what everyone in Silicon Valley was thinking: Google should have completely dominated the AI race from day one

Google invented the Transformer architecture in 2017. Then, every single one of the eight authors on that seminal paper left.

Noam Shazeer left to start Character-AI. Aidan Gomez founded Cohere. Ashish Vaswani and Niki Parmar started Essential AI. Jakob Uszkoreit founded Inceptive. Llion Jones founded Sakana AI. Lukasz Kaiser joined OpenAI. Illia Polosukhin built NEAR Protocol.

Google had the paper, the compute, the users, and the distribution. What they lacked was a culture that allowed the people who invented the tech to actually ship it.

The talent didn't leave for higher salaries. They left because bureaucracy slowed down execution. Within eighteen months of leaving, every single author was shipping groundbreaking models somewhere else.

The lesson goes far beyond Google. It’s what happens to any incumbent when protecting legacy revenue becomes more important than shipping the future.

Sundar Pichai kept the CEO chair. The eight engineers who wrote the paper built the rest of the AI industry.

Follow if you'd rather build for the future than protect the past.

[Freddy @0x_freddy]

@akamr7up base model capabilities have limits. agent orchestration layer is pure architectural necessity, not cash-grab bloat

[Freddy @0x_freddy]

Andrew Ng invented the online AI education industry. His 2011 Stanford ML course pulled 100,000 students overnight and became the template every $500 AI course has copied since

Last week he walked back into Stanford and gave the industry he built six months to live.

His words: "Prompting will be dead in 6 months. Graphs are what's replacing it."

Ng co-founded Coursera. Ran Google Brain. Was Chief Scientist at Baidu. He personally trained more AI engineers than any university on earth. When he names a deadline, the deadline is real.

The whole $500 "learn prompt engineering" niche was already downstream of his old course. Everyone was teaching a distilled version of what he taught in 2011.

Now he's showing them what to teach next. Two hours at Stanford. Graphs, loops, autonomous agents. The first ten minutes cover what every current course sells at $500 a seat.

The man who made this a $2 billion industry just walked in and turned the lights off.

Six months from now, half the coaches selling "prompt engineering" will have quietly pivoted to "graph engineering." The other half will be looking for a job.

[Anatoli Kopadze @AnatoliKopadze]

https://t.co/fyBGWQbsvn

[Freddy @0x_freddy]

@Kwadjo_Derrick fair point on capital, but capital without execution culture gets you slow corporate inertia. look at google's compute budget vs their launch timeline

[Freddy @0x_freddy]

Sam Altman admitted what everyone in Silicon Valley was thinking: Google should have completely dominated the AI race from day one

Google invented the Transformer architecture in 2017. Then, every single one of the eight authors on that seminal paper left.

Noam Shazeer left to start Character-AI. Aidan Gomez founded Cohere. Ashish Vaswani and Niki Parmar started Essential AI. Jakob Uszkoreit founded Inceptive. Llion Jones founded Sakana AI. Lukasz Kaiser joined OpenAI. Illia Polosukhin built NEAR Protocol.

Google had the paper, the compute, the users, and the distribution. What they lacked was a culture that allowed the people who invented the tech to actually ship it.

The talent didn't leave for higher salaries. They left because bureaucracy slowed down execution. Within eighteen months of leaving, every single author was shipping groundbreaking models somewhere else.

The lesson goes far beyond Google. It’s what happens to any incumbent when protecting legacy revenue becomes more important than shipping the future.

Sundar Pichai kept the CEO chair. The eight engineers who wrote the paper built the rest of the AI industry.

Follow if you'd rather build for the future than protect the past.

[Freddy @0x_freddy]

Columbia charges $80,000 a year to teach diversification. In 2002, a hedge fund manager sat in a classroom upstairs and taught the opposite. His method had returned 50% annually for ten straight years

His name is Joel Greenblatt. He ran Gotham Capital from 1985 to 1995. He held just a few positions at a time. Almost no one in finance sustains 50% for a single year. He did it for a decade, returned all capital to outside investors, and walked into a classroom to give the framework away.

The core of his lecture isn't just about formulas. It's about finding structural market inefficiencies — spinoffs, restructurings, and forced-selling events where institutional investors dump stocks regardless of price. He explains why these opportunities persist even after everyone knows they exist.

Every stock screener is free now. Every filing is searchable. The bottleneck was never access to information. It was knowing which information to ignore.

Filmed from the back row. Audio uneven. Students blocking the frame. He handed a 50% annual playbook to a room of 30 people for free.

Watch it before your broker sends you another 60-stock "balanced portfolio."

[Freddy @0x_freddy]

@rpjb exact. spun-out ventures backed by google compute would have given them equity in https://t.co/gsE3PvPO26, cohere, and sakana from day 1

[Freddy @0x_freddy]

@nickhATX true, but when holding back a breakthrough to protect legacy products lets 8 authors walk out, it’s organizational failure

[Freddy @0x_freddy]

A math problem stood unsolved for 50 years. A grad student cracked it in less than a week and was at MIT fourteen months later

Lisa Piccirillo, 2018. She heard about the Conway knot at a topology conference and started poking at it in the evenings. Her own words: she wouldn't work on it during the day because she didn't think it was "real math."

The problem had been open since 1970. Every knot with twelve or fewer crossings had its "sliceness" figured out - except one. John Conway's knot.

She built a different knot that shared the same four-dimensional trace as Conway's, then proved hers wasn't slice. Since trace-siblings share slice status, Conway's wasn't slice either.

She mentioned the proof casually to Cameron Gordon, a senior topologist at UT Austin. He started shouting, "Why aren't you more excited?" and pushed her to submit it to the Annals of Mathematics.

The field itself has one bizarre property: Every dimension has a single smooth structure. Except dimension four. Not three, not five, not eleven. Four has uncountably infinite smooth structures. Nobody fully knows why.

Topologists debated it for half a century. She solved it as an evening project.

Follow if you'd rather read about real mathematical breakthroughs than corporate buzzwords.

[@kleos @1kleos1]

https://t.co/2qhV8sPbLd

[Freddy @0x_freddy]

@steveky1er 2 hours of foundational architecture beats 200 hours of superficial prompt tutorials every single time

[Freddy @0x_freddy]

Stanford just uploaded a 2-hour course covering the exact knowledge Anthropic pays $750,000 a year for. It builds an LLM from scratch

Not the marketing tour. The engineering: how tokenization actually splits language, how the model decodes a prompt, what happens inside the training pipeline, how the architecture fits together end to end.

- 00:03 Breaking Down Tokenization Mechanics

- 25:47 Inside the Prompt Decoding Process

- 35:43 The Complete Model Training Pipeline

- 1:16:50 Building the Core LLM Architecture

Two years of self-study bootcamps have tried to teach this. Stanford put the definitive version on YouTube for free. Two hours from the people who write the papers you're reading beats two years of tutorials from people who read them second-hand.

Most engineers use LLMs every day and can't explain what happens between their prompt and the answer. The people getting hired at $750K can.

Follow if you'd rather understand the model than keep guessing at it.

[Roan @RohOnChain]

https://t.co/8HHG7saiut

[Freddy @0x_freddy]

Columbia charges $80,000 a year to teach diversification. In 2002, a hedge fund manager sat in a classroom upstairs and taught the opposite. His method had returned 50% annually for ten straight years

His name is Joel Greenblatt. He ran Gotham Capital from 1985 to 1995. He held just a few positions at a time. Almost no one in finance sustains 50% for a single year. He did it for a decade, returned all capital to outside investors, and walked into a classroom to give the framework away.

The core of his lecture isn't just about formulas. It's about finding structural market inefficiencies — spinoffs, restructurings, and forced-selling events where institutional investors dump stocks regardless of price. He explains why these opportunities persist even after everyone knows they exist.

Every stock screener is free now. Every filing is searchable. The bottleneck was never access to information. It was knowing which information to ignore.

Filmed from the back row. Audio uneven. Students blocking the frame. He handed a 50% annual playbook to a room of 30 people for free.

Watch it before your broker sends you another 60-stock "balanced portfolio."

[Tigerflow @tigerfl0w]

https://t.co/Ye9wCJ9Jiv

[Freddy @0x_freddy]

Lee Kuan Yew took Singapore from a third-world island to a first-world power in one generation. In 2000, he stood at Harvard and spent 2.5 hours explaining exactly how

No PowerPoint. No sound bites. No prepared quotes ready to be screenshot. Just a man who had done it, talking to people who wanted to.

He covered how to pick people who tell you the truth. How to decide fast without pretending the answer was obvious. How to hold a nation together when everyone around you says it's impossible. Why every leader eventually chooses between being liked and being effective.

Two decades of Harvard MBAs have said the same thing about it: 2.5 hours from someone who actually built a country beats two years of case studies about it.

It's on YouTube. It has aged better than almost every leadership book published since.

Skip Netflix tonight. This is what authority sounds like when it stops trying to impress.

Follow if you'd rather learn leadership from the people who built things than from the people who wrote about them

[Freddy @0x_freddy]

A guy shipped a playable HTML5 shooter yesterday for 40 cents in tokens. He used Claude 3.5 and DeepSeek R1 inside Cursor. Last month he cleared $4,200 doing this on repeat

You move, aim, shoot. Enemies pathfind. Generated code, WebGL graphics, live URL in 4 hours. Token spend: less than a cheap coffee.

His system: one prototype every two days. Ships it to itch-io and web portals with rewarded ads baked in. Most get 50 plays. Every 10th hits a trend and pays for the other nine.

Ad networks, tips, and micro-sponsorships added up to $4,200 last month. Tool bill: $25. Everything else is margin.

No 5-year indie dev suffering. No 100-page design doc. Prototype, ship, watch metrics, kill or scale.

Indie studios still debate art styles on Discord. He typed a prompt before lunch and had a live link by dinner.

The question isn't whether AI can write games. It's how many iterations you can ship before competitors wake up.

40 cents in. Live game out. Multiply by every mechanic you haven't tested yet.

Comment "CURSOR" for his rules and setup. Follow for the next teardown

[Freddy @0x_freddy]

@0xSizer exactly

[Freddy @0x_freddy]

A guy typed one prompt and got a cel-shaded boat racing game he now sells on Steam for $12,000 a month. Total build cost: $423

690 million tokens. No engine, no team, no year of iteration. He described the game he wanted, Claude wrote it, he shipped it.

The game looks like a lost Wave Race sequel. Anime aesthetic, cel-shaded water, drift meter, lap times, four AI opponents chasing you across a cranes course on the open ocean. It runs, it feels good, it sold.

He didn't beat AAA studios. He beat every other solo dev to the finish line by two years. Studios still write design docs. He wrote a prompt.

The game industry took a decade to make indie possible. Claude took one prompt to make solo shipping trivial.

The next hits on Steam won't come from studios with 200 people. They will come from whoever describes the game they wish existed and hits enter before anyone else does.

$423 in, $12,000 out. The 28x return isn't the interesting part. The interesting part is how many other prompts nobody has typed yet.

[Freddy @0x_freddy]

A 21-year-old in Brazil made $7,632 last month running a Fanvue for a woman who does not exist. She is an AI in a hard hat.

764 men pay $9.99 a month to watch her tear down walls.

Build cost: $200 in Claude credits. Time from idea to first paycheck: 37 days.

He never touched a camera. Claude generated every video, every caption, every reply. He picked the niche and the outfit. Everything else was a prompt.

The construction girl posts on TikTok. Sweat, dust, safety goggles, tight leggings. She never speaks. She never breaks character. Men follow her for the aesthetic, then click through to her Fanvue for the version they actually paid for.

The market for AI bikini girls is saturated. The market for AI girls in work boots was empty.

The lesson is not that AI is the moat. The lesson is that the moat is knowing what men have not seen this week.

[Cip⚡️ @cipgerx]

https://t.co/pgreHQRyL8

[Freddy @0x_freddy]

@CryptoMavka 3k a month from 4 channels is like 800 bucks per channel before taxes. is that really worth the conveyor setup