The Era of Personalized Software?

A few years ago when I went on my first run with a baby stroller, I realized that my stats were off. I’ve long been an avid user of an app called Strava to track my runs and connect with other runners in my community. It’s a fantastic app. But when running with a stroller, I was a bit slower, heart rate higher, this specific activity was a blip in my stats.

I thought, surely there must be a “running with stroller” feature in such a widespread app? Nope. And still isn’t, despite many folks requesting it.

I get it. It’s a niche requirement, and Strava couldn’t justify complicating things or slow down their progress for the majority of their audience by catering to every specific need. This is product management 101 and I understand this perspective all too well. But in 2024 things are different, and the economics of producing software are changing; have LLMs and the boom in AI made software personalization a near-term reality? Let’s break it down.

As of today, August 25th, 2024, you can ask Claude to write basic web apps for you, and it will do so. Software written just for you based on your requirements. Not only that, but you can also ask for guidance on deployment, basically making what you have just created accessible to others as well.

For something like my above example, my engineering brain screams the bulleted complexities, but I’m actively trying to reprogram my thinking here and challenge this. Why couldn’t you have a version built just for you or a niche community, built off a fork of the main code line, that does exactly what you want it to do? There could be a crowdsourced set of specific features automagically pooled and delivered by an army of AI agents that can constantly keep up with patches and updates to the main line, and perform all the CI/CD necessary to make your version available to you.

We can now use LLM software engineers to re-architect existing software platforms to the point where you have a personalized app that can be adjusted for features that are specific to you. Depending on the unit economics, this could be extended to small communities of like minded users (i.e. stroller running parents), where this personalized version could be discussed within that community, none of whom are engineers, and evolve that variant as they see fit? Tools could guide these communities towards reasonable product-management-in-the-age-of-personalized-software practices. And then at the outset, you end up with an app that is personalized for you, can even have personalized community features, but doesn’t necessarily slow down the main development.

For this to be possible, you have to reckon with the complexities of open source development. This includes challenges like version control, forking, maintenance, reviews, malware, and the trade-offs of these decisions. But there is a gargantuan trail of data here, and surely LLMs can either guide us down the best-known path and probably in not too long emerge something new that works for them better than it ever has for us.

I believe (and hope!) software personalization is on the horizon, and it opens up a whole new door for platform as a service (PaaS) vendors that can ship and deploy these personalized variants to the given app stores, clouds, etc. Deploying software (i.e. DevOps and CI/CD and the act of taking coded ideas and making them available in a browser or on your phone) is one aspect that is incredibly well suited to LLMs, is very rarely a differentiator for businesses, and non-IT teams should never have to toil with. Vercel is playing in all the camps here right now, and the more they can weave the model companies like OpenAI with their stacks and deployments, the further they’ll climb towards a major player in PaaS at large.

At Paloma, we’ve taken steps towards software personalization in a small way by providing our customers with a highly customizable event intake process that is AI supported. It’s not necessarily doing anything groundbreaking at the moment, but this design has been at the back of my mind from the start, with the intention of expanding and exploring its potential in the future and with a variety of verticals. Lots to play with here, and the initial shaping of it has already seen promise during development, and for our customers.

Additional Thoughts

A couple of additional things to throw out here. As open source becomes more prevalent, I believe the future is file, a la File over App. This philosophy which prioritizes interoperable files over data-on-the-inside apps, is a precursor to all the above. That’s the Obsidian future, an app I absolutely adore for note taking (and personal knowledge management). Walled gardens are going to be out, democratic data is in (see OpenMined and this reddit announcement for more of where I see this going). Using apps that keep your data locked inside without easily interoperating, creates risk and makes me hesitant to invest in it’s use.

Personalized software will likely require users to have full ownership of their data, stored in highly interoperable formats. Everybody needs access to their data after the app has moved on and been forked into oblivion and overtaken by something else, period. That’s it.

Economics of all this? Ya, that’ll be tough. We’ll figure it out! We as in, the market will figure it out. Many of the new AI companies are putting their code out there as open source and letting folks that are keen self host it, and simply charging for its use on Cloud — this is a great model, and one we’ll continue to see more of.

I feel the era of personalized software is near and we’re on the cusp of a revolution in how we customize our digital tools. It’s an exciting time, and I can’t wait to see what innovations and possibilities emerge as we explore this new frontier.

Anybody know of any examples of personalized software in the wild right now? Via AI, or other means? Micro or macro scale? I’d love to hear about it if so. Also if you think I’m totally blowing smoke, lay it on me!!

(full disclosure; I wrote all of this, and Claude Opus helped me make it easier for y’all to digest (hopefully??). Thanks for reading!!)