Be More Ambitious, Fail More

Couldn’t agree more with Ethan here.

My general advice for using AI for any purpose is to ask yourself: "how do I make what I am doing even more ambitious?"

You do this until you either hit a wall and then you know the shape of the frontier (useful!) or you find out that the AI can do more than you think (common!)

— Ethan Mollick (@emollick) October 25, 2026

Coincidentally, I was having a similar conversation with a friend recently. I told him he had absolutely no idea how far AI models had come, and that not being ambitious in 2026, and not letting that ambition expand as the models get better, is to commit the sin of sloth.

I see it in my own case too. I’m attempting things I have no business even thinking about, let alone dreaming of. A few examples: thisindianlife.today, measure.thisindianlife.today, tipsheet.markets, and fromthedumpsterfire.com (coming soon; it should be live in a few days, still a WIP). And even now, in 2026, I keep getting surprised by just how capable these models are. Yes, the capabilities are jagged. There are peaks and troughs. But I’m still surprised by the ridiculous amount of capability I’ve yet to unearth.

A lot of people are worried about what we lose as we rely on these models. Most of those debates can be summarized under the umbrella of ceding cognitive autonomy. My tentative view is that maybe that’s not a useful frame. I’m working on a post arguing that these models aren’t just giving us superpowers. They’re giving us cognitive freedom. They give us the freedom to attempt things we otherwise wouldn’t, and to learn things we couldn’t have done, had no access to, or wouldn’t have dreamt of.

In the best sense of the phrase, this is the golden age of having fun and fucking around. And the menu of things you can fuck around and find out with keeps expanding every day. Even if model capabilities were frozen at today’s level, the benefits you could still get from them are insane.

I’m not saying this from the point of view that everything will work out. But as Ethan says, as you keep hitting walls, you get a better sense of the shape of the jaggedness. What these models can do, and what they can’t.

By hitting walls, I mean you should actively want to fail more. Take extra-ambitious shots to figure out what’s possible and what isn’t. Every failure is valuable feedback that you can fold into your mental model of how these models work, and of what they’re capable of and what they aren’t. That mental model is always tentative, given how fast the models keep improving. But there’s a whole lot to learn from hitting walls and failing outright. At least, that’s been my experience with these tools.

But in all the fucking around, the failing, the hitting walls and the doing, perhaps the most valuable thing has been discovering for myself what I like and what I don’t, what I enjoy, what I have fun with, and, most importantly, what makes me whole. I’ve noticed that I’m having more fun after AI than I did before it. I’m doing more things, and in the process, a lot of wonderful things have been happening. I’ve got to speak to and meet people I otherwise never would have. I’ve discovered new interests, and rediscovered old ones, that give me that warm, gooey feeling inside. Hitting walls has opened up doors and windows in places where there used to be only walls.

I heard Josh Woodward, who runs Google Labs and the Gemini app, describe something like this on Lenny Rachitsky’s podcast. His team keeps a running list of what’s almost possible, and when something on that list crosses over, they get a team around it and just go:

I mean, one thing we try to do a lot, and I think this would apply in any of those situations, is how do you get yourself to the frontier? And that can be both with playing with the models, but also just watching what are the types of experiences that have just become possible where we have a phrase on the team, we talk about what’s almost possible and we just have a list actually where we’re trying to track those things.

And when one crosses over, it’s almost like a phase shift when something goes from solid to gas or liquid to gas or something, you’re like, “Ooh, that just became possible. I saw one of these two days ago and I can’t stop thinking about it.”

It’s the same with me. Some of the things I’ve built were ideas I’d been sitting on for a while, but in the early days the models were simply too weak to pull them off. Now they’ll pretty much do it in a day or two.

So yeah, be more ambitious. The advantage of being ambitious right now is that it’s still very, very early days for LLM adoption. I’m often shocked at how many smart, capable people aren’t using these tools. I’m convinced that if they did, their smarts and capabilities would be magnified to such a degree that the world would be better off for it.

I don’t say this from a position of AI boosterism. I say it from simple, cold, hardcore utilitarian logic. These are tools that enhance your capabilities and let you do more. In a way, you can use them to create clones of yourself. By that I mean you can now have multiple agents working simultaneously, doing all the things that would otherwise rot on your to-do list.

I’ve also heard from people doing this that there’s a cognitive cost to it. They’re losing their flow state, or they’re getting drained much earlier in the day.

This is a new shift, a new way of working, and it’s still very, very early days. We’re all groping in the dark trying to figure out if this is good or bad, the best of times or the worst of times, or whether we’ll all crash and burn. Personally, I’m pushing myself as much as possible to figure out where my limits are.

It’s become a cliché to say the human is now the bottleneck, but there’s some truth to it. These models are indefatigable. They can keep going on and on and on. But me, my consciousness stuck in this miserable container of flesh and bone, has to do the disgusting thing of getting some sleep. How I wish there was a pill that cured sleep so I could vibe code 24/7. I kid. Sort of.

That being said, I wouldn’t be surprised if there are negatives, and often large negatives, to this style of working. I’m not saying one should easily dismiss losing your flow state. But the point I’m trying to make is that the old ways of working are dead. The new way is having a fleet, a swarm of agents at your beck and call, doing the things you want them to do.

Which goes back to a point I’ve made multiple times in earlier posts: using AI tools is a trade-off. There’s a famous Thomas Sowell quote, “There are no solutions, only trade-offs.” I’ve used it a hundred times, and I keep it as a mental prompt in my head for thinking about all the choices I make with these tools.

Go forth and be ambitious. If Jesus were here today, I think he’d have said something to that effect.

Pair this with Felix, Dan and Khe

Pair this post with How to be more ambitious: a tutorial for humans with models by Felix Rieseberg. It covers a lot of the same ground, and it’s a beautifully written post.

Hi, I'm Felix. I'm an eng lead working on Claude. I love watching people take flight and do things they never thought possible.

We often describe that as "being more ambitious" when using AI, which is tricky advice to follow. I wrote a tutorial:

felixrieseberg.com/how-to-be-more…

— Felix Rieseberg (@felixrieseberg) October 13, 2026

One passage I loved:

Accomplish hard things by believing them to be possible: In 1939, George Dantzig showed up late to a statistics class at Berkeley, copied two problems off the blackboard, and handed in his solutions thinking they were homework. They were two unsolved problems in statistics.

Accomplish hard things after watching somebody else do them: For decades, humans chased the four-minute mile. Roger Bannister broke it in May 1954 but remained the only human to do so for a mere 46 days. If you’re into running, we’re watching something similar happen to the two-hour marathon: Sabastian Sawe ran the first official sub-two-hour marathon in London this April. In the same race, the runner-up also broke the two-hour mark.

Accomplish hard things by giving yourself permission to fail: Apparently everyone’s favorite business book is The Inner Game of Tennis, which I discovered just last week. The book’s core thesis is that you’re not playing your best game of tennis if you’re trying really, really hard. We engineers might remember Linux, announced as “just a hobby, won’t be big and professional like gnu”.

And just as I was about to publish this post, I came across this post by Dan Luu, who says something similar:

BTW, when people have said they used the new model and everything just works, software is solved, etc., I haven’t found this to be true at all. It’s true that there are a lot more things the newer model can one-shot, where things basically just work, but this just means that I end up working on more ambitious things where models still struggle. I’m more productive and can work faster, but I find that, in terms of what I’m doing, I get farther from “it just works” with each more capable model because the gap between what the model can verify effectively without significant handholding and what the model can implement has been increasing. All of that is a long-winded and less eloquent way of saying: we should be more ambitious with the software we build.

Is this the universe sending me a sign that I’m right? Can’t be a coincidence. It’s a coincidence, I know. I’m kidding.

And then it happened again. I’m not making this up: I opened Substack, and the first post on my screen was Are you being ambitious enough? by Khe Hy. Khe starts from a feeling I suspect a lot of people share:

“You need to be more ambitious with AI.” This advice regularly gets dispensed by engineers at the frontier labs. And I hate it. Not only does it make me feel permanently behind. But it reveals my own shortcomings.

Then he builds on Felix’s essay and turns it into something tangible: a skill file you can download to make any task or prompt more ambitious.

Okay, now I’m convinced. The universe is telling me to be more ambitious. It’s time to move past things that merely induce AI psychosis and on to outright AI madness.

I publish these posts on my newsletter as well.

---

A caveat: this is a stream of consciousness written after seeing Ethan’s tweet. I know there’s a lot of vagueness in a lot of what I’ve said here. I’m not writing this because it’s a final, solidly frozen thought. It’s a way of probing my own views and staking out a thinking space to explore all of these ideas: on work, on doing more, on being more ambitious and having an experimental mindset, on what you lose and what you gain when you use AI, on the trade-off between cognitive autonomy and freedom, on what losing the old ways of working means for us and what the new way means, and a bunch of other things.

These are still very early days in the age of AI. As the shadow of AI continues to expand, I’m personally excited to explore all of these questions and what they mean for me. And something tells me all of us will be forced to do the same.

So to be clear, I don’t write this from a place of certainty, but from a place of wanting to explore the deeper questions. Feedback, ideas, thoughts welcome.