|
|
Tell me if you’ve been here before. |
You ask your AI to build one small tool. Something basic, like a way to help you ship blog posts faster, or maybe something that helps with your email. |
Twenty minutes later, it works… And that's exactly when the trouble starts. |
Because now you're looking at it thinking: this could use a settings page. And dark mode. And an export button in case you ever want the data elsewhere. |
Each of these features is effortlessly coded in minutes by your trusted AI model of choice. So you keep building more and more features on top of what was once a simple, single-purpose app for a single user — you. |
I've been building a lot of things with AI, and I've noticed the hardest part is never getting the thing to work. It's being content once it does. |
Adding a feature has never been this cheap, which means the old excuse for not adding it (too much work) is gone. But there are hidden costs involved with mindless expansion of features. Just because it's cheaper for you to add features doesn't mean it's cheaper for your customers to absorb the complexity and cognitive load.
Here’s a concrete example: I’ve been working into the wee hours every night on a new product, YouSpot (The AI-native CRM for 1 person companies -- more details at the bottom of this article). It launched in beta 6 days ago and it’s been going really well. I’ve been thinking about building this product for 20 years. I have a ton of ideas. And that’s the problem. Although I’ve been thinking about ideas that would make for a 10 year roadmap, YouSpot customers haven’t. So, I spend as much time curating the ideas -- and even more importantly, for features I do add to the product, I spend the extra time to figure out how I can keep the overall system cohesive. It’s like building a house when bricks are almost free. You don’t want to just slap up a 6 story tall structure with 20,000 square feet. You want to be mindful of the foundation. Determine what is needed and be a careful curator of ideas. |
The tendency to FeatureMaxx leads me to believe that, in the age of AI, fortune favors the focused. |
Today I want to break down: |
Why a feature's real cost was never the cost of building it (the FeatureMaxx trap) How AI raises the stakes on focused, intentional building (and how you can adjust) The trait that separates the people who build businesses from the people who endlessly improve things
|
—@dharmesh |
|
|
The Cost of Adding New Features |
When people talk about the cost of a feature, they usually mean the cost of building it. That’s true. But it’s also first-order thinking. |
The way I learned to think about it at HubSpot is that a feature has costs in orders. |
The first order is implementation: what it takes to design and code the thing. |
The second order is maintenance: what it costs to keep it alive, month after month, as everything around it keeps changing. |
Then there's a third order, which is the biggest one and the one almost everyone forgets. It's what the feature costs your customer. Not once, but every single time they use your product. |
Every extra button, every extra setting, every extra path through the interface is a little more complexity they have to hold in their head, and a little more cognitive load they carry forever. |
Just because it's cheaper for you to add features doesn't mean it's cheaper for your customers to absorb the complexity and cognitive load. |
AI collapsed the first cost -- but it’s important to be mindful that it’s done almost nothing to the third. |
Being able to spin up a working tool in an afternoon is one of the best things to happen to builders. But cheaper and faster only expands what's possible to build. You're still not going to ship something people rely on over the long haul without real skill and real judgment behind it. |
The building got cheaper and faster. The expert judgment required to create well-designed software didn't get any easier. |
|
|
Why Focused Products Are Suddenly Harder to Make |
Complexity is the default direction of pretty much everything. It’s the second law of thermodynamics which paraphrased is “left alone, most things will deteriorate into random, chaos.” |
I've said for years that simplicity is worth fighting for, and that it requires fighting for, because the universe is working against you. |
Left alone, every product gets more complicated. Nobody decides to make it that way -- features and settings just pile up, because that's the natural order of things. |
So the rule I try to keep top of mind is simple: when in doubt, take the simpler path. There's always time to make something more complicated later. It's always harder to make it simple again. |
Complexity doesn't just annoy people. It kills companies. Maybe not as quickly as running out of money, but much more reliably. |
Back when building was expensive, the cost of building acted as a natural brake. You couldn't really FeatureMaxx even if you wanted to, because you didn't have the time or the engineers to pull it off. |
AI removed the brake. The one force that used to keep products focused, sheer difficulty, is basically gone. Which means the focus now has to come from you. |
Naturally, you can use AI itself to help you police this. |
🤖 Try using a prompt like this to help determine whether a new feature is worth adding or not: |
The one job this tool is supposed to do is: [job]. The current feature list is: [list]. For each feature, tell me what it costs the user in complexity and cognitive load every single time they use the product, not what it cost me to build, and argue for whether it should be cut. |
|
|
|
|
Who Are You Building For? |
AI hands the same leverage to two kinds of builders: |
One ships something narrow, puts it in front of real people, and lets what those people actually do decide what gets built next. The other keeps improving in private. The tool grows a dashboard, a settings page, an export button, and never once meets a user.
|
I'll be honest that the second one is a trap I fall into as easily as anyone. The pull to keep polishing something nobody has used yet is stronger than ever these days. |
It’s why I did this weird thing with YouSpot. The product is currently 6 days old and in beta. But I put a page up so customers (we already have 1,600+ of them) could submit ideas for YouSpot (you don’t have to be a customer to see the ideas — they’re public). And other customers could upvote/downvote them. The top idea right now already has 143 votes! I haven’t promised that I’m going to implement an idea just because it has a lot of votes but I’m definitely going to be more judicious about building other random features that are not on the list. I’ll do it — I’ll just be more thoughtful about it. |
Companies and individual builders alike need to avoid constantly shifting the goalposts. In the early years at HubSpot we even made this mechanical for a while: one feature in meant one feature out. It was a very coarse measurement, but a coarse constraint beats no constraint at all. |
It also helps to remember who you’re building for. Building for yourself? Go crazy -- add as many features as you’d like. You’re the target audience after all. This kind of building is an excellent way for beginners to learn more about AI by building with AI. |
But if you’re trying to build a business, this kind of endless building can become a distraction. As you know, there’s much more to business than adding new features. |
Before you build another feature, write down the one sentence that says what the thing you're making is actually for. Just one sentence, the single job it exists to do. Then build until it does that job, and ship it right there. |
The building has never been cheaper, which is exactly what makes saying no worth more than it used to be. I’ll say it again: Fortune favors the focused. |
If you give YouSpot a try, hit reply and tell me what you think -- I'm shipping daily, so feedback doesn't just get read, it often gets built 🙂 |
—Dharmesh (@dharmesh)
P.S. YouSpot is the AI-native solo CRM I recently launched as part of HubSpot Next. The tl;dr: it's built for one-person businesses -- founders, consultants, freelancers, creators -- not for GTM teams (that's what HubSpot is for) and not for consumers. You connect the tools you already use (Gmail, calendar, X/twitter, LinkedIn, etc) and it builds a "second brain" from your business context, then surfaces the relationships, follow-ups, and opportunities that deserve your attention. You chat with YouSpot the same way you'd chat with ChatGPT or Claude. In fact, it gives you access to all the frontier models in one place. You should try out Claude Fable 5.1 in YouSpot (it launched yesterday). The product is still early but getting better (but not always bigger) every day. It’s easy to get started. Just $10/month -- cancel anytime. |
|
|
|
|
What'd you think of today's email?Click below to let me know. |
|
|
