The new era of software

AI lowered the barrier to building. Software supply exploded. Distribution, trust, taste, and judgment matter more now.

Building is no longer the bottleneck

For most of software history, building was the bottleneck. If you had an app idea and weren't technical, your options were pretty limited.

OptionReality
Hire developersExpensive
Find a technical cofounderBasically startup dating
Learn to codeYears
Give up on the ideaSurprisingly common

That was the moat for a long time: access to execution. Five years ago, building anything real meant a small team, decent budget, and three months. Now one person can open a tool like Lovable, Claude or Replit and generate a real UI, auth, a database, onboarding, a half-working product, and at least one feature or dashboard nobody asked for. A long lunch with a side of extra confidence, and you have an app.

Which means a lot of people are building apps right now who couldn't before (like me). Marketers with strong eye. Designers who would have needed a technical cofounder. And regular people who have had an idea for years and finally have a way to prototype it.

The bottleneck moved. Now the hard part is everything around the build, like, you know, running the actual business.

The software supply explosion

The internet is filling up with software extremely fast right now. New apps, agents, generators, internal tools, Chrome extensions, you name it. Categories with a handful of competitors a year ago now have dozens. And the bar to ship the next one keeps getting lower.

The software supply explosion

Most of what's online right now was shipped in the last eighteen months.

People who joined GitHub in 2025
36M
Repositories created in 2025
121M
AI projects live — nearly 2× since 2023
4.3M
LLM SDK usage, year over year
+178%
New Shopify apps in a single week
700+

Most of that growth is recent. Most of what's online right now was shipped in the last eighteen months. The Shopify App Store is the version of this you can actually watch happen in real time. It's the kind of number people don't believe until they look it up.

Being new doesn't get you noticed anymore. Neither does being functional. Both used to be impressive on their own and now they're table stakes. The products that pull ahead are the ones doing something the other forty in their category aren't.

"Good" is no longer enough

Before vibe coding came around, users were willing to put up with "bad" UI if no other alternative solutions existed. Now they look at Claude.

Today, when I'm evaluating an app or tool and I hit friction, my first thought isn't "what else is out there" anymore. It's:

Here's the thing, I actually don't WANT to build it myself. I just want a solution that works, is enjoyable to look at, isn't bloated with 100 features I don't need, doesn't ask me 15 onboarding questions, and doesn't cost an arm and a leg. Is that too much to ask?

And it's not just "build it myself" anymore. The "good enough" alternative used to be spreadsheets and duct-taped workflows. Now it's ChatGPT, Claude, a Notion doc, Zapier, Airtable, or some other chaotic workflow somebody built at 1:14am that somehow still works.

The DIY option is quietly breaking a lot of products' moats. Polished UI used to signal effort, skill, resources, legitimacy. Now it mostly signals someone had access to AI. A polished sidebar has convinced a lot of people they had product-market fit, when what they actually had was generated screens.

Polished UI isn't a differentiator anymore.

Products increasingly compete on the things that are harder to automate:

  • Clarity — knowing what your product is actually for
  • Positioning — being the obvious pick for a specific kind of person
  • Trust — earned over time, not declared on a landing page
  • Reliability — the boring stuff that keeps people from churning
  • Distribution — how you reach the people you're for
  • Taste — knowing what to leave out

None of those come out of a build prompt. They come from someone who picked a specific person and built like that person was the only one who mattered:

ProductPositioned againstWhat they actually won on
LinearJiraBeing for people who notice keyboard shortcuts
TallyTypeformBeing for Notion users who didn't want to pay $50/month for a form
BeehiivSubstackBeing for operators who treat newsletters as a growth channel
GranolaOtterBeing for people who actually take notes in meetings, not just record them
SuperhumanGmailBeing for people who would pay to make email feel fast
CronGoogle CalendarBeing for people who care about how their calendar looks

None of these positioned themselves as a "better" version of an existing tool. They built sharper, not broader.

Most products won't. Which is why a lot of them are going to struggle in the middle. Not bad enough to fail immediately, not useful enough to become essential. Just kind of… there.

The question users are quietly asking when they land on a new product isn't "can this team build software." That's assumed. It's "is there a reason this exists, other than the fact that someone could ship it."

The valuable part isn't execution anymore

Taste sounds vague, but in practice it's mostly the discipline of subtraction — what you choose to build, what you leave out, and what you refuse to add when someone requests a hyper-specific feature you can already tell only they want.

AI tools are extremely good at adding things. The harder skill is deciding what's actually worth building. Which is why the rare thing now isn't the ability to ship. It's making good decisions consistently, understanding users deeply enough to know which problems are real, and operating something long enough that people start to depend on it.

Judgment, taste, and understanding your users are skills that compound the more you practice them. They're what the next generation of products will be built on.