Weekly writing about how technology and people intersect. By day, I’m building Daybreak to partner with early-stage founders. By night, I’m writing Digital Native about market trends and startup opportunities.
If you haven’t subscribed, join 70,000+ weekly readers by subscribing here:
The Personalization of Software
There’s a quote from Dmitry Shapiro, Myspace’s former CTO, that I often think about:
When you go out to a bar, you don’t put on a white and blue uniform, a Facebook uniform. You’d put on all sorts of amazing things to stand out. You’d want to be radically different from all others. And that was the real value of Myspace. When people talk about missing Myspace, they miss customizing their profile.
Before Facebook, social media was messy and chaotic and personal. MySpace was a free-for-all for self-expression. Then Facebook came along and put you in a blue-and-white straitjacket, constraining personalization.
The MySpace → Facebook shift meant more conformity, and Shapiro’s comment above gets at people’s desire for customization and individuality. As the internet has grown, we’ve seen a similar shift to conformity happen in many categories. This includes software: with more scale, products gravitate toward mass-market.
Yes, yes—there’s more software than ever before. We’re drowning in apps. Apple filed a trademark for “There’s an app for that” way back in 2009.
And yes, you could argue that more software = more personalization. But most of us have phones full of apps built for the common denominator, not necessarily built for us. With more users, technology applications actually get less tailored for our needs.
AI is changing this: as the cost of building software drops precipitously, software is getting personal. People want customization and individuality when using software (just like they want customization and individuality in social expression), and AI makes personalized software newly possible.
In January, we wrote about Lovable in The Age of Fragmentation in Content and Code.
Back then, Lovable was just a couple million in ARR. They’ve come a long way since: the chart here comes from Bond Capital’s excellent AI report, while Lovable’s Anton Osika tweeted on Sunday that Lovable had hit the $60M mark.
Lovable dramatically compresses time from idea to execution.
Seven months back, a Reddit user challenged himself to make 30 apps in 30 days using Lovable (the product was much worse back then, by the way). He had some successes, including a Chrome extension that summarizes YouTube videos while you watch, which got 1,000+ users in its first week.
Over the weekend, I saw a cool story on LinkedIn. I can’t find it now, but someone posted that he’d been shoe shopping over the weekend. While his wife tried on shoes, the shoe store attendant asked the man what he does for a living. “Tech,” the man said. The attendant explained he’d had this idea for an app, but didn’t have the skills or the time to build it. The man asked the attendant if he’d seen Lovable—the attendant hadn’t—and then they built a working prototype of the attendant’s app idea in under 10 minutes, all while the wife was still trying on shoes!
This stuff is really cool; I recommend playing around with products like Lovable. Anyone, with no technical skills, can build some pretty solid stuff.
Say I want to build an app that lets me track what I eat; I put in this prompt:
Within about 10 seconds, I’ve got my Meal Tracker app:
I can hit “Publish” and voila—you can even use it here: https://meal-muse-journal.lovable.app/
This entire flow—from idea to shipped product—took under 30 seconds.
You’ll notice that the link for my meal tracker is a Lovable link: lovable.app. One thing that’s interesting: people are making stuff for themselves, but not publishing nearly as much. Anish Acharya pointed out that web traffic to lovable.dev is 3x higher than traffic to lovable.app:
So far, custom-built software isn’t really meant for others; it’s meant for our own personal use, to solve our own personal problems.
My opinion: this will eventually change. Right now, products like Lovable and Bolt are still in the “early adopter” phase, meaning that they’re largely used by tech-forward people. But most people aren’t like that; most of my friends probably won’t be building their own apps. I agree with Olivia that most people don’t have the initiative (let alone ideas!) to build their own apps. Rather, long term we’ll see more people consuming customized, personalized apps; more builders, yes, but also many, many more users of hyper-personalized apps.
Okay, one other topic: there’s a lot of software out there, but it’s…kind of bad?
“Software ate the world, but it sucks”
This is the opening sentence to a manifesto written by Zach Witzel, the co-founder and CEO of Helium. Helium is one of our Daybreak portfolio companies, and its vision is to build “self-improving software.” What does that mean?
Zach’s builds on his spicy opener:
In 2011, Marc Andreessen wrote the now-famous “Software is Eating the World” post, and oh boy was he right. Since he wrote that post, global software expenditure has risen 760% to $1.24 trillion, with the software-heavy Magnificent 7 now making up more than 30% of the S&P 500.
AI has unleashed a new tsunami of personalization and problem solving, but most applications we use today are still rigid, static, and plainly…sucky.
Before starting Helium, Zach was at Uber, where he led much of the experimentation org. Experimentation is a fancy way of saying: testing what works, then doubling down on it. Experimenting on a piece of copy in Uber’s app could easily take a month, if not more. The typical workflow would look something like this:
Brainstorm: Identify of a bunch of ideas to improve the Uber experience
Stack-rank: Dive into the data (or user research) to stack-rank which ones to actually work on
Define: PM writes a spec of the top option
Design: Designer explores potential versions, working with the PM to align on the right experience
🛑 Delay: Wait for the engineers to free up
Develop: Implement the new experience
Test: Walk through test steps to confirm no degredations
Configure: Set up and launch the new experiment
🛑 Delay: Wait weeks for usage data
⚠️ Forget: Probably forget it’s running. Come back months later
Assess: Look into the results, identify the winner
Finalize: Go back to engineers to ship the winner
…then do it over again
Helium’s vision is to automate this process by building software that continuously improves: Helium's agents work day and night to improve software applications, continuously identifying opportunities, then building out new features to improve user experiences or lift business metrics.
The company is starting with paywalls, a thorny problem for any mobile subscription business. Mobile was the birthplace of product experimentation and growth—led by Zynga and mobile gaming—and it’s an enormous market full of PMs, engineers, and growth folks wanting to pull their hair out.
By automating paywall testing, Helium can raise revenue (often by >50%). The roadmap then includes onboarding flows, messages, price changes, discounts, and so on.
(If you’re a mobile subscription company, you can email Zach at zach@tryhelium.com and try out Helium.)
Software was “invented” in 1948: that was the first time a stored-program computer held a piece of software in electronic memory and executed it successfully, at the University of Manchester in England. Three-quarters of a century later, software is more flexible and more personal and more self-driving.
Satya Nadella recently said that 30% of Microsoft’s code is written by AI. Products like Cursor have completely reimagined coding. Products like Lovable, meanwhile, are rapidly changed what non-technical people can build—how software can be like Play-Doh, shaped and maneuvered into new forms.
And companies like Helium are picks and shovels that give software agency to iterate and improve. Not every company has Uber’s product org or Duolingo’s growth team; Helium’s vision is to make software build and iterate and refine itself, making an AI the rockstar product leader in your company.
My favorite chart from the Bond AI report was this one:
It nicely captures compounding—exponential growth.
How can you look at this chart and not be a fan of technology? Sure, tech should have guardrails and be balanced by good policy. But it’s also the engine behind economic growth. The world is a lot richer now than in the past (the distribution of that wealth is another problem; see: prior comment on policy!), and AI is the next force multiplier.
There are 8.1B people in the world; an estimated 25-50M can code software. That’s about 0.3-0.6% of the population. So far, making stuff with technology building blocks has been limited to that group. That’s now changing.
Thanks for reading! Subscribe here to receive Digital Native in your inbox each week: