I was recently pairing on a round of DX interviews with our team's PM. A detail that really struck me was how many folks had created custom software to support their work. It was a broad mix of agent harness helpers, custom tooling to spin up environments, custom software to gather and work items (pull requests, issues, etc), custom to-do trackers, and other productivity tooling. All sorts of software just for themselves.

I've always done a little bit of this, as I'm sure most developers do. But the cost before was so high that it had to really solve a problem for it to be worthwhile. Native Mac apps, for example, haven't been a thing I've ever had a sufficient mix of motivation (to learn Swift) and need to build something pre-LLM. The cost of building something for myself has now plummeted so low, that, like everyone we interviewed, I similarly have a larger collection of software that solves a problem just for me. Most of mine haven't targeted personal productivity, though I plan to think what I could make in that space after hearing what so many other devs have built for themselves.

Broadly, this seems to really be a thing. SDT's The Headless Lifestyle included a good discussion around using agents to create a personalized UI into complex consoles or other enterprise apps. I personally don't buy the saaspocalypse thesis. Enterprises need solid, centralized store for data. But I do think that overly complex GUIs that serve dozens of different personas will be something many users abandon as they move to personalized or fully headless interactions. Just give us clearly documented API or CLI. Never having to interact with Workday's generic-to-the-point-of-incomprehensible UI would be amazing. DoorDash even just released a CLI.

I'm also seeing this from technical folks who aren't developers. The Verge discussed their vibe coding experience the other day. Platformer did an interview with one of the founders of Raycast, who thinks this is actually a whole new product category and is building Glaze for that reason. I'm less sanguine about that, I doubt folks need a specialized layer on top of the existing products for this, but the discussion is interesting in that it hits the same themes. His view lines up with mine that personalized, single user or team apps are really taking off. And in a business context, personalized view into shared repositories of data will be a big part of how work gets done.

I read this the other day, which I think is really relevant to any discussion around LLMs and writing software:

A developer vibe-coding a side project a dozen people will ever run, and a team keeping a ten-year-old enterprise system alive for another quarter, share almost no constraints worth naming, and most of the advice in circulation is really one of those two people telling the other how to live.

I do both these! But personalized software is very much the former. It can be as polished or janky as you need it. It can have known bugs that just don't bother you. Really, the only thing that matters is that it solves a problem for you.

My own growing list of personal software is less work focused, though now that I have some additional ideas I plan to revisit that. The details of each would be very boring to someone else (I generally don't even use public repos, since I figure no-one else cares about my vibe-coded stuff). But each of these is software that solved an issue for me, and is something I use weekly if not daily. Broadly, I have

• A Mac menu bar app that reads my calendar and then displays a modal and plays a sound one minute before a meeting
• A tool to scrape and parse websites into an RSS feed (each site is parsed and served into a separate feed)
• A really simple Audio Unit host for a single instrument and an effect chain, so I can try out plugins and make presets without opening something complex like a DAW or Gig Performer
• A health website that aggregates workout Garmin and Withings data. Existing tools suffer the same issue as enterprise software, they try to serve too many different personas. I actually know multiple people who have built themselves something similar
• A small, text-based sequencer that takes input for notes, chords, duration, and velocity (like c4 * 0.25, g4 * 0.25, (c4 e4 g4) * 0.5 @ 100) and outputs MIDI
• A OneTab clone with syncing. The developers redid the UI and I hate it, so I cloned the original and also added a Google Drive sync.
• An app that records MIDI of all my piano practice and organizes it into session files, so I have fully history of progress and can also listen back to improve
• An iPad app similar to iReal Pro with a lot more instruments, a different system to input the chords, and more customizable patterns
• An progressive webapp to drill key signatures

Those are the ones I'm actually using and have polished. I've tried and discarded a great many more since the friction to experiment is now so low.

The future of how exactly we use LLMs in large projects and collaborating across teams isn't certain to me. They're obviously going to be a part of software development going forward, but exactly how is changing almost monthly. But personalized software is something that's a lot clearer to me. Personal software is the area where I've had the most fun. I'm making myself tools that I get tons of use from; things I previously just didn't have the time and energy to build. Simple to medium complexity software, written mostly or completely by an LLM for a single person, is very much a thing. Personalized software, both for work and personal use, really has arrived.