10 days. That's all it took. A team of just four people at Anthropic shipped a full feature. They called it co -work. Four people in 10 days. And you have to contrast that with what's happening right now in most companies, where teams are probably spending 30 days just to make a roadmap. Just to plan the pilot? Just to plan it. It's wild. It really is staggering. It feels like two completely different worlds operating at the same time. I think they are. And the gap.
It's not talent. It's not about who has the smarter engineers. No. It's a mindset gap. It's all about what's expensive and what's cheap in, you know, 2026. Welcome back to The Deep Dive. It is Sunday, January 25th, 2026. And if you're feeling a bit overwhelmed professionally right now, I think it's safe to say you're not alone. It feels chaotic. It really does. But we're looking at this amazing piece from Max Ann today. And his whole argument is that the chaos, it's not random. It's a symptom.
He says the bottleneck of work has moved. Exactly. For decades, the hard part was doing the thing, the execution. Now, that's the easy part. And that one little shift is breaking. Well, everything. So today, that's what we're going to unpack. We'll look at why execution is basically a commodity now, what the new constraints are, because they're not what you think. Right. Not coding or writing. And then we'll walk through these eight what he calls toxic rituals that we all need to break.
Sounds good. I want to start with this idea of the inversion of cost, because I think for most of us, we still have this. this trauma almost from how slow and expensive projects used to be. Oh, absolutely. I mean, for the last, what, 30 years? Execution was the most expensive thing you could buy. People's time, engineering hours. It was also precious. Right. And so we built all these processes around it, like walls to protect this very expensive resource. We had
to. It was like a bank vault. You don't just leave the door open, right? So we invented the PRD, the product requirement document. We had approval gates, endless planning meetings. And that all made sense. It made perfect sense. If it takes a developer... It can show it's the right one. But now, here we are. 2026. And that whole ratio has just completely flipped. AI made execution cheap, almost free, almost instant. The source brings up Cursor, the AI code editor.
I think most people have heard of them. For sure. But look at their growth. They went from $1 million to $500 million in revenue faster than any company in Sauce history. And the way they work, launching a huge new product isn't some big stressful thing. It's just. It's just another Tuesday. Wow. That's wild. Because execution isn't their bottleneck. And this gets to a really core concept, the manufacturing principle. Okay, let's pause on that because
it's so important. The article says bottlenecks don't disappear, they just move. Can you break that down? Sure. Think of an old car factory. The slowest part of the line, the bottleneck, is attaching the wheels. It takes three people 20 minutes. Okay, so you've got cars piling up, waiting for wheels. Exactly. Now you invent a robot that puts wheels on instantly. Snap. Done. Problem solved. Well, you solved that problem. But does the factory now make infinite cars?
No, because now they're all piling up waiting for, I don't know, the engine to be installed. Exactly. The bottleneck moves. That's the whole theory of constraints. You fix one thing and it just reveals the next slowest part of the system. So for the last 20 years. The wheel attaching was the coding itself. It was. That's what took all the time and money. And now AI has basically given everyone an instant wheel attaching robot.
So the bottleneck has just shifted violently to what comes before the code and what comes after. So if building the thing is no longer the hard part, where did the difficulty go? It moved to clarity and distribution, knowing exactly what to build and getting it in front of people. Okay. So the value is on the what, not the how. That's the entire game now. So let's dig into that. Max Am points to four new constraints. The first one is clarity. Right. He calls this
the billion dollar question. When you can build anything with a single prompt, the danger isn't that you can't build it. It's that you'll build the wrong thing. Very, very fast. He says that writing a PRD now often costs more in terms of time than just. building a prototype. Now, I have to say, I can hear a thousand product managers just yelling right now. You can't just stop writing requirements, can you? That's chaos. I think the argument is more about the format of the
requirement. A 10 -page document is just words. It's squishy. It's open to interpretation. But if I can prompt an AI to build a rough version of the interface in 10 minutes, That's not open to interpretation. The prototype is the requirement. The code is the spec. So the document is now the waste. The document is the expensive part. Yeah. Which leads right into the second constraint he mentions. Ambition. This one's more psychological. When shipping was hard, you got maybe three or
four big swings a year. So you had to be careful. Yeah. You played it safe. But if you can shift something every 10 days. That's 50 swings a year. The risk isn't failing anymore. The risk is just timidity. It's thinking too small. He uses that horseless carriage analogy. Yeah, it's a classic. When cars first showed up, people just saw them as carriages without the horse. They didn't really reimagine what transportation could be. And a lot of us are doing that with AI right now. Totally.
I mean, I have to admit, I mostly use it to write my emails a little faster. Me too. We're just putting a motor on the same old workflow. We're not asking the bigger question, which is, why are we even sending emails at all? Okay, third constraint, distribution. This one is huge because if anyone can build a product, the product itself isn't your competitive advantage anymore. It's not the moat. And the example is Cognition, who made Devon. They partnered with Infosys. Which
seems so weird at first, right? This cool, cutting -edge AI company partnering with this giant kind of traditional consulting firm. Because Infosys has access. They have a 300 ,000 -person sales and distribution network. Exactly. It's a signal that the value has flipped. The magic box AI model isn't the rare asset anymore. The human network is. The relationships. So distribution is the new moat. The code is just the water. Perfect way to put it. Which brings up the last
constraint. Relationships. He has this great line. You can't vibe code trust. I love that. It's so true. Tools change. Platforms change. The only thing that lasts is human trust. AI can't automate that. It sounds like we are shifting from engineering problems to human problems, doesn't it? Technology became easy, so judgment and connection became the premium assets. Okay, so the world has changed. Yeah. But most of us are still working with habits from, you know,
2015. Toxic rituals. Toxic rituals. He lists eight of them. And I have to admit, this part of the article made me squirm. Because I looked at this list and realized, I'll probably do three of these this morning. It's very uncomfortable reading, yeah. Let's do the first four. They're all about action over permission. The first one is the permission loop. Ah, yes. The old check a line, get buy -in dance. Which, again, made
sense when a mistake was super expensive. But now, asking for permission often takes longer than just doing the thing. He mentions a tool called Manus. Yeah, what was that? It's a tool that builds a presentation for you while you are in the meeting talking about it. Whoa. So by the time you've finished asking for permission to build the deck, the deck is already done. That just highlights the absurdity of the old way. But okay, let me play devil's advocate.
If everyone just defaults to doing, doesn't that just create total chaos? It might, but the cost of that chaos is low. That's the key. If I build the wrong thing in an afternoon, we just throw it away. Who cares? If I spend three weeks in meetings getting permission to build something and then it's wrong, that's a huge waste. So ask for forgiveness, not permission. That's the new rule, which connects to habit number two. Polish as procrastination. Oh, this is a tough
one for perfectionists. The old 80 -20 rule. Spending 80 % of your time on that last 20 % of Polish. And that's a waste of time. It's a defense mechanism. It's hiding. The new advice is ship ugly. Look at Notebook LM from Google. It wasn't perfect when it launched, but they got it out there, got feedback, and made it better in public. Okay. Habit three is one I think everyone will feel. meetings as default. The silent killer of productivity. A one -hour meeting with six
people isn't one hour. It's six hours of execution time gone forever. The fix he suggests is to replace meetings with demos. Yes. Code is a way of getting ideas into contact with reality. Instead of arguing about a theoretical feature, just build a rough version and show it to people. You can't argue with something that works. Number four, structured waiting. We used to think waiting was, you know, a neutral part of the job. You send an email, you wait. And now? Now waiting
is a blocker. If you're blocked, you make a provisional decision and you just keep moving. You assume the answer is yes because the cost of being wrong and fixing it later is now lower than the cost of waiting. This feels uncomfortable for people who pride themselves on being careful, right? It is, but caution has become a tax on speed safety now comes from iterating, not planning. Well, speaking of planning, let's get to the next four habits. Number five is inverted planning
and doing. And I have to say, I love a good plan. I love a roadmap. I know. It feels so responsible, doesn't it? It feels like you're doing work. But this flips that on its head. We used to think planning was cheap and safe and doing was expensive and risky. Now it's the other way around. Planning is the luxury. It's expensive because our predictions are almost always wrong. Execution, though, is cheap, and it gives you real, actual data. So
his advice is to cut planning by 90%. 90%. I know it sounds insane, but the idea is to let reality inform your plan, not the other way around. Build something, see what happens, and then make your plan. Which connects to habit number six, the deck instead of the demo. Slides are just too slow. They hide the truth behind bullet points. A working prototype answers questions way faster than a 30 -slide deck. We hide behind decks. It feels safer. Safer for our egos, maybe? Not
for the business. Okay. Habit 7. Consensus before action. This is that fake alignment we all know. You're in a meeting. Everyone nods and says yes. And then nothing happens. Exactly. He says, I tried X, and here's what happened. Is way more persuasive than, let's all agree to try X. Results create consensus. Arguments don't. And the last one, number eight, hoarding until ready. This is the big ego check. We hide our work because we don't want to look stupid. We don't want to
show something that's broken. But showing half -finished work saves you from what he calls quiet mistakes. I like that term. It's great. A quiet mistake is when you spend a month building something perfectly, only to find out nobody wanted it. If you'd just shown the ugly, broken version on day three. Someone would have told you. So the logic is that showing something broken today is better than showing something perfect next month. Yes, because by next month, the market
or the tech has already moved on. We've covered a lot of ground here. I want to address the skeptics when we come back, because I know someone listening works in, say, health care and is thinking this all sounds really reckless. We'll be right back. We're back. So we've been talking about breaking all these old habits, shipping ugly, skipping meetings, acting without permission. But I have to represent the person listening who's a surgeon
or a compliance officer at a bank. Right. They're thinking, if I ship ugly, people could die or I could go to jail. And that's a completely fair point. The argument isn't to use AI to, you know, hallucinate a bridge design and just start building it. So this doesn't apply to high stakes fields. I think it does. The challenge is to really ask yourself. How much of our process is actually required for safety or by law? And how much is just tradition? How much is just the way we've
always done things? Right. Is that meeting a legal requirement or is it just a habit? Exactly. You can have compliance without a 30 -day roadmap. The document isn't the safety. The actual safety protocols are the safety. And this all comes back to velocity. It really does. The tools are basically equal now. Everyone has the same AI models. You have Claude, I have Claude. So the only real differentiator left is velocity. And it's not just about being fast for the sake of
being fast. No, that's such a crucial point. It's not about speed. It's about learning. Velocity multiplies intelligence. It just compresses your learning cycles. So if you can iterate... 10 times while your competitor iterates once. You're going to learn 10 times faster. The winners are going to look like Anthropic and Cursor. They'll be shipping while everyone else is still in a planning meeting. So is this just about moving
fast or is it about learning fast? It's learning speed allows you to fail small so you can succeed big before anyone else starts. OK, so let's try and bring this all home. We started out saying that 2026 feels chaotic. And the argument is that the chaos is just friction. It's the friction of using old habits in a totally new world. It's like trying to put a horseshoe on a Tesla. Right. It doesn't work anymore. The big takeaway then is that execution is no longer the hard part.
The bottleneck moved to clarity, knowing what to build, and distribution, getting it seen. And to deal with that, we have to shift our whole mindset from permission and planning to default to doing. The PRD is dead. The prototype is king. It's a scary thought, though. Yeah. It means giving up all that busy work that feels so safe. It does. And that's the thought I really want to leave you with. All those meetings, all the slide decks, the endless emails, they feel productive.
They feel like work. But are they? Or are they just a comfort blanket that's protecting you from the scary reality of actually shipping something? That is the question. So here's the challenge for this week. Look at your calendar. Find that next alignment meeting. You know the one I'm talking about. We all have one. Cancel it. And just use that hour to build the ugliest, roughest version of the idea you were going to talk about. See what happens. Thanks for diving in with us. Until next time.