This is nice, but Apple essentially blocks sideloading (outside test flight environment) and Google is moving there as well. While it has never been easier to produce individualized software, it was never more difficult on the official platforms, unless one wants to create their own OS (at some point that might be possible as well).
* A personal music player for Apple Music with an embedded agent loop, so I can say things like "I'm going to work in the wood shop, play woodworking-appropriate music and keep the playlist full and non-repeating indefinitely".
* A calorie/protein tracker that takes 1-liner summaries of meals, resolves them to macros with GPT5 calls, and tracks them in SQLite.
* A macOS menubar combined Apple TV and Roku TV remote control.
* A calculator front-end to SageMath that renders Latex math notation and has point/click shortcut affordances for linear algebra and MVC.
* A "self driving wiki" macOS desktop llmwiki that imports and digests PDFs.
* A menubar tracker for the little temperature sensors I have scattered all throughout the house so I can click my menubar and instantly see the temps in the different rooms of the house.
* A layout tool for designing picture frames to get cut lists and rabbet depths and stuff.
That's just the desktop stuff. These applications are generally better than what I'd get from the App Store. The calorie tracker is one of the better-designed apps on my dock.
I feel like I'm not doing this enough yet, it hasn't fully sunk in, and I'm going to be building a lot more stuff when it does sink in. I haven't yet gotten to the point of writing my own note-taking application (still Bear.app for that), but how long can that last?
> I feel like I'm not doing this enough yet, it hasn't fully sunk in, and I'm going to be building a lot more stuff when it does sink in. I haven't yet gotten to the point of writing my own note-taking application (still Bear.app for that), but how long can that last?
Indeed.
I'm not sure when, but at some point after I saw Claude Code, I realised what I wanted for my next OS was basically an LCARS-themed* thing, where the "desktop" is a TUI and whenever I needed an app I didn't already have, I'd ask an LLM to write it for me instead of downloading it (benefit: the app would never have an update pending (which might break my flow) that I had to install before use, unless I'd specifically asked for a new feature to be written); and if I wanted to browse something on the web, the browser would use something like an LLM to strip out everything that wasn't content.
LLMs aren't quite good enough for that yet. They're interesting, I've also made a bunch of apps this way that was an easier experience than even figuring out which offering on the app store was right for me**, but still, the LLMs are not quite good enough; I'm pining for the OS equivalent of a car with a self-driving mode so good there's no steering wheel and the windows have black-out shutters so you can sleep while it drives you and your insurance is fine with this.
As a new father, just using so many apps sounds exhausting, let alone building and maintaining them. I would have been very excited about these maybe 10-15 years ago. Now I just want the ability to do things with my phone when my hands are tied up holding the baby.
Tell me about it. Holding a 4mo old right now, nap trapped :-) Speaking of which - Claude Code wrote me a web proxy for HN that rewrites CSS for the dark mode - precisely for this situation. Written and deployed with Claude Code remote control too, also while nap trapped.
> A personal music player for Apple Music with an embedded agent loop, so I can say things like "I'm going to work in the wood shop, play woodworking-appropriate music and keep the playlist full and non-repeating indefinitely".
how is AI (or a human for that matter) supposed to know what “woodworking-appropriate music is”? I had the same thoughts when spotify released their “let AI make a playlist for you”. Like, we already have recommender algorithms, why do we need to get a LLM involved?Maybe I just don’t understand woodworking and the connection it has to music though (I’m also the kind of person who’s had the same playlist for like the past 4 years)
But, I do want to say, personal app built just the way you like it, an alternative front end to a closed source ecosystem even, is super cool, and has been one of the things that AI coding has really enabled.
I’ve done a few project where I just give claude a har file and tell it to reverse engineer a cli from some web interface and it’s so powerful to be able to do stuff like that now
Guaranteed it would be unable to avoid searching for and selecting songs that specifically mention wood, woodworking, the shop, etc. It's like when you tell it to write something and give it a "vibe" or "mood" word wishing it could just give tone to the generated text, but instead it incorporates it literally.
Because I played a bunch of music while doing woodworking stuff and asked for more like that. Things like this are easy with a simple agent loop. It's not picking random music out of the sky; it's working from my Apple Music library (and then expanding on it from the Apple Music catalog).
>I played a bunch of music while doing woodworking stuff and asked for more like that.
I would probably think about doing a little workspace sharing and ask for more live musicians to come in and do some serious woodshedding themselves ;)
I know some that are getting that buzz-saw sound out of their instruments fairly repetitively already, and there's no place to go but up :)
> It’s not like “play more music like this” is something we ever needed AI for.
Borderline off-topic, but I love how fast the goal post moves. A decade or two ago "Similar music to this" was definitively in the realm of AI/ML, now it's some full-stack dev's afternoon to implement.
I don‘t think someone disputes that this is in the realm of AI/ML. I just think its questionable if this kind of usecase is genuinely something were an LLM is necessary or outshines the more traditional approaches
What are you even talking about? No, of course music recommendation systems had nothing to do with LLMs two decades ago (which is obviously the crux of the discussion), that is insane. And no one has said anything about a “full-stack dev's afternoon”, I’ll kindly ask that you don’t straw man. No goal posts are moving, that fallacy does not mean “argument I don’t like or understand”. What happened to good faith discussion? It’s ever more common on HN that someone will just invent some argument in their head, attack that, then act like they are the arbiter of reason when they couldn’t even understand the argument. That’s not interesting. Engage with what was said.
The true power of LLM powered interfaces is the malleability of the whole thing. Sure you can write traditional software to handle “more like this”, but how hard would it be to handle “more like this but uptempo”, “more like this but not the same artist”, ad infinitum?
I have a bunch of skills Ive created for random work that have “predefined interfaces” like a single url. I very frequently trigger these skills with freetext prompts instead and get all the intelligence of the skill while bypassing what it was actually designed for. This aspect of malleable API is extremely powerful and is probably the only reason I keep AI in the loop of some of my custom software vs removing it and avoiding the cost/nondeterminism.
Recommendation algorithms are, frankly, weird as a user. Something watches your tiny interactions and, with zero concept of anything happening, gets surprisingly good at predicting your next interactions.
It’d be like eating and having an alien with no concept of food or taste staring at how much your facial muscles change and making notes. Then recommending places to you based on the other people they’ve been studying.
> how is AI (or a human for that matter) supposed to know what “woodworking-appropriate music is”?
Well someone is going to be concentrating, but it’s a long period, with loud activity. maybe you’d guess at decently active music, not relying on careful complex listening, probably fairly steady rather than sudden dramatic changes.
Maybe it could save previous playlists and what you’ve said about them, or ask questions like “so like, really calm or techno or what?”. Adding just a bit of common sense makes an enormous difference with this kind of thing.
A typical (in fact, the modal) line from the database, which goes back to April, is "oyakadon 1/2c rice, 2 thighs, 1/2 breast, 2 eggs, bunch broccolini, std oyakadon broth", and the model works out for itself that these are average-sized thighs and breast (boneless and skinless, because it's oyakadon). "std oyakadon broth" is a reference to a "house recipe", of which there's a side database, each costed by GPT5. There's a low/med/high adjuster on every entry; sometimes the thighs are bigger or smaller, for instance.
I've:
* Checked these against all three major frontier models and gotten the same macros breakdown within relatively tight bands
* Manually researched a couple of the most common recipes
* Run this system over many months against my most likely TDEE and reconciled against scale weight loss (which has been profound).
My confidence in frontier model's ability to generate "good enough, probably more than good enough" from this level of meal description is quite high.
Key simplifying assumption (for the model): I'm cooking most of my meals. I cook. I virtually never order out or eat packaged food of any sort. This would be a harder problem and the answers less reliable if I was giving it, like, "1/2 JP Graziano muffaletta, extra hot giardiniera".
You can actually get just as good results for packaged food if you just allow the agent to use a web search tool & keep a growing local database of "known food items". I did the same for a while, but it was quite expensive (this was a while back with less powerful models and required a bit of orchestration to make it reliable)
For basically anything but chicken I give exact gram weights of proteins and of any fats or carbs. I generally don't give it any precision about vegetables that won't drive protein, fat, or carbs. Most meal components aren't going to meaningfully change the macro breakdown except for fiber, which I don't care enough about to track.
This is going to be a funny thread to show my doctor. I think something that isn't getting through to people is: I'm not taking The Internet's advice on this stuff; I hired a person to handle these questions for me. Highly recommended!
I quite literally, right at this moment, am sitting next to a friend who is a doctor with a focus on nutrition and they are saying that whoever told you that fiber doesn’t matter is going against almost all of the modern understanding of health science and research.
His advice is literally “He needs to just Google it.”
I feel on the opposite side of a divide when I see these kinds of comments here about people creating bespoke software to address deficiencies they perceive in their daily lives. There's an engineer mindset that tools can solve any and every problem, but do they? Are you quantifiably any fitter than you were before? Better bloodwork, BMI, body fat percentage, anything? Are you creating better wooden goods? More wooden goods? If you track mood in any way, are you at least consistently happier while working because of the auto-generated playlists? Have you gotten any better at math thanks to the pdfs?
When I reflect on my own life and think about why I'm not personally adopting LLMs for anything, it occurs to me that I just don't feel like I have any problems like this to address. I was already eating appropriately, already working out appropriately, already happy with the music I listen to, already happy with the way I learn and practice math. I could make a billion personalized apps, but I was already a two-time state champion in cross-country and track who has never been overweight, got perfect scores on the SAT and GRE, won regional academic olympics in math, all before mobile apps existed and before I ever owned a personal computer of any kind. That's not to say I can't do even better than I did as a teenager in the 90s and I try to maintain some proficiency in these things into middle age, but it never occurs to me that I need or should need software to help me with this. I have no clue what the temperature is right now of any room in my house, but I know they're all in a comfortable range that isn't hurting any machinery or animal and that seems good enough to me.
I see all these enthusiastic booster posts and comments about all the tools you can make, but I don't feel like what is missing from my life is I don't have enough tools. I wasn't sitting around prior to GPT-3 thinking to myself "you know what would improve my life? More software." There are deficiencies and problems to address for sure, but it's mostly issues of prioritization and motivation. I have fallen off the wagon a bit this year and don't exercise nearly as much as I should. I don't talk to my wife enough. I don't talk to other friends at all. I don't go outside enough. But those aren't problems to address with software, either. If anything, if I spent more time making software, even with LLMs, that would be more time not going outside, talking to my wife, or working out. No matter how much raw output I could get computers to output, more time spent fiddling with computers is not going to improve my life. I shouldn't even be here right now, reading and commenting on Hacker News. The sun will be up and it will be 90 degrees fahrenheit within an hour of that. This is my last chance of the day to go run without feeling extremely uncomfortable while doing so.
This isn't a "oh my God look at me I'm so much better than you" comment or at least that's not what I'm trying to get across. But I'd encourage the original poster here and you as well to be outcome-oriented. My weight isn't in a warning zone. This guy is posting stats, so I am leaner than he is and my marathon time is faster, all without the aid of any tracking apps. You're a public figure, so I know you're in the same time zone as me, which means you were posting your own comment at midnight last night, midnight on a Friday night reading Hacker News. My sleep isn't perfect, but I'm not reading Hacker News at midnight on a Friday. Are all these apps actually making your life better?
> There's an engineer mindset that tools can solve any and every problem, but do they?
No. These people just spend more time fiddling with their computers instead of doing the parts of life that it supposedly frees them up to enjoy. It's IoT lightswitch automations all over again. There's no reason anyone would ever need to calculate their smoothie according to their Strava data and the positions of the planets on that day. It makes far more sense to spend more time knowing your body and learning to treat it well than trying to create an app that will do it for you.
As someone fiddling with computers trying to solve "every problem," the only actual solution for me is making tools to assist with being aware of the state of your body. Overly simple example, but give someone a heartrate monitor for the first time and they won't be able to guess their current bpm all that accurately. If they pay a little attention to it for a month, you'll get much more closer guesses for instantaneous rate. If it becomes easier to be aware of your current state, you get a more accurate model, and can start to predict basic cycles etc. Eventually, without the tools.
I am wildly, drastically healthier today than I was when I started this in April. This is something I should have been tracking and measuring a long time ago.
Don't think it has to be, if you are mindful of it. Ai uses fewer dependencies usually, as it doesn't mind rewriting things, and you can easily add scans and sanity checks to your sdlc
It's somewhat unfortunate that programming is still pretty hard to start with.
Early home computers would boot up with BASIC. Granted, the user experience was horrible, but at least you could start programming right away.
What is stopping us from creating an operating system that has programming as a first class citizen?
There is JavaScript in the browser, but that still feels as an afterthought, and incurs all kinds of security problems.
Why must we wait for AI to work around this? Is it really true that only very few people are interested in programming? Why not teach it in primary school, along with reading and writing?
I joked a while back that it was faster to vibe-code a white noise generator than to download a white-noise app.
I've since tested that theory on the Google Play Store (iOS has one built in), and it was depressingly true. The slow part was trying to find an app that didn't waste your time with ads.
There's one in particular that I use quite often and have for about a year, vibed for myself: it's a chat interface that walks you through processing an emotional or difficult moment, following a process / workflow. Supports Martin Seligman's ABCDE and Byron Katie's "The Work". Two techniques that I found most useful to improve my thinking and responses to difficult situations. You converse with it, and it leads you through the stages of the selected flow. More than a system prompt - it tracks and progresses through the workflow at the right times, and you download a consistent pdf of key details from the "session" at the end for your records.
The thing is, I can't sell it. I don't even think I can open source it. It's mental health (minefield of legal and ethical issues), and would probably breach copyright, trademark etc law.
Yet it's very useful to me. I guess it's the equivalent of making a system to do these things at home, like prompt cards or key points taken from the books, with a well formatted note taking structure. Except it's way easier because you just chat like you're talking to someone.
I would never have spent time building this. But now it's so cheap and easy to do.
Side note, why not just deploy onto a VPS and save the money from using hosted services, as well as protecting yourself from when they inevitably go down? The CBA clearly shows that while active development is going on its trivial to have hosted service subscriptions because the AI sub is way more, but what if that sub disappears? You're left paying (relatively) exorbitant amounts compared to $1-10 per month per VPS for life, or less if you have the privilege and wherewithal to self host. It's also dead easy with AI. Maybe this just reinforces how much a personal runtime solution is needed, as stated in YC's recent RFS.
Yeah this is what I did. All my personal services are running on one EC2 instance and I use subdomains to separate them. If I need more memory it's easy to scale up. I just use sqlite on the instance for my database. You don't need a postgres service if it's just for you.
Hosted services are only great for setup and scaling. You're not going to need scaling. Infra setup can now be done easily with AI and terraform.
That's what I did after I realized how much I'd pay for managed databases and auth.
I now have 3 VPS on Hetzner that run about $14/month in total (the old pricing, thankfully). I mostly got three so I wouldn't worry about wiping one if something terrible happens.
Works like a charm. Even with the new pricing, Hetzner or DO would be cheaper than what OP uses. And the maintenance isn't that hard.
Easier to leave vulnerabilities but you can also do it on Vercel and Neon.
Almost $2000 a year is beyond ridiculous and would make me question the author's ability to write software on his own (but I'd get downvoted for insulting him or not realizing some kind of time savings they help with).
I spend $2/month on my VPS, write my code in HTML/CSS/Javascript with Posgresql along with other Unix (FreeBSD) tools and the standard languages and get similar things done.
Tbf if I don’t enjoy that aspect and it saves me a day or so per year that I can use for consulting instead it makes sense to do that trade. Two and a half perhaps including taxes.
Yeah Vercels an awful service for stuff like this, paying $20/mo may not seem like a lot but its $15/mo or $180/year more than you need to be paying.
It still baffles me why the heck people pay for services line neon and supabase as well. Utter madness imo. If you add that on to something to get rid of OP could've saved $540 per year.
Thariq wrote "personal software was a bit early in 2020 but in 2026, it really can be as personal as a home cooked meal, or a handwritten letter."
Excuse me??? Vibe-coding personal software is not a homecooked meal or a handwritten letter! It is [at best!] more akin to tastefully combining microwaved ramen with microwaved veggies and toaster oven chicken, or using ChatGPT to write a letter then editing it afterwards. It's fine, maybe even very good. But it's not "homecooked."
Ugh. The most depressing thing about the AI boom is watching tech people devalue human experience.
If someone spends an hour in the kitchen making you food, and they choose to use the microwave to heat up the veggies because it's healthier than boiling them on the stove, that's devaluing the human experience? What's more human than listening to another person's problems and then making something to help them with that problem? It could be a chair or a meal, or now, a bit of software.
> Every so often, you get a glimpse into a whole world you never knew existed.
I had no idea that people did this. This sounds expensive.
Just like other people have no idea that we (Scandinavia) let babies sleep outside in a baby carriage wrapped in a tiny sleeping bag. Fresh air is awesome. Cold air is even better.
This works for some babies and is not universal in any way.
There is no limit really. The baby is in warm clothes and in a sleeping bag during winter. Less clothes when its warm.
My children slept outside while it was below zero C. The only time they were not sleeping outside was while they were ill.. having a cold or something.
My partner is a sleep consultant, and most of the time the families she works with are at their wits end and their marriage is hanging by a thread. Having young kids is incredibly hard on people and relationships.
If you want to optimise and have the money to spare, I think it's fine. There are going to be people in the world that are better than you at most given tasks, and if working on those tasks is not fun for you and delegation is not a net negative, why not? It's why personal trainers exist (though I don't like them for separate reasons)
Its clearly a net negative, though, because missing a nap is now causing tremendous anxiety to the point where they'd rather spend a week chatting with Claude to build an app which handles the consequences of missed naps than just spending some time with their infant, maybe reading to them so they fall asleep.
In Precious Little Sleep the writer says some half of babies sleep well and some half require help. We must have had the former and so there’s no advice to give. If you were in that set it’s possible that your actions changed something or it’s possible it’s just what it is.
Have you raised kids? This is the least shocking expenditure I've heard of; if this had existed when my kids were born I might have shelled out for it.
Let me share an anecdote from my soviet childhood.
Like most soviet children of the 80s, I was outsourced to my kindergarden each mom's workday.
I abolutely loathed everything about it, but the most fascist thing to my infant mind was the post-afternoon nap.
For the life of me, I just couldn't make myself fall asleep, and had to pretend, making extraordinary efforts to really look like sleeping.
We were literally scanned crib by crib, aisle by aisle by the surveilling nanny to detect sleep pretenders.
Those offenders of Morpheus who were caught red-handed, were routinely ordered to step out in front of the entire dorm, which of course was instantly awake and gleefully staring through eyelashes now, and do 20 humiliating squats.
Needless to say, that resulted in nothing but fierce hatred towards the soviet system, which treated children as future soldiers, and oppresion in general.
So let me disagree with you on day napping being a fundamentally capitalist thing, haha :)
The only difference is that it plants the seeds of hatred towards different entities - either the system or the parents.
I would have thought so too until our third kid was impossible to out down.
We managed the first two so we are not completely incompetent:) so no it turned out very useful to have a "sleep coach" give us strategies that we hadn't thought of.
Nothing as elaborate as what's above but not completely ridiculous.
To me it sounded like they're paying someone to plan not only their kids' naps but also their own AND spend their spare time writing an app to help keep up with the recommended nap schedule. Just take a fucking nap, maybe after going outside for a walk in the park.
Gosh, going outside for a walk in the park doesn't buy conversion for AI companies, that's such a luddite thing to propose to someone who's got skin in the game.
No, it’s a plague on anxious, over-optimised middle class parents.
The idea that someone can send over a schedule in advance is pure voodoo. Much like huckleberry, the ml app that makes new parents insane over wake windows.
The examples in this thread are all personal-life software (music, calories, thermostats). What I keep wondering about is the other end: the boring business admin that self-employed people carry — invoicing, chasing late payers, expense categorisation, tax set-asides, quoting. That's "software for one" too, except the stakes are money and most people seem to limp along with a spreadsheet plus a shoebox of receipts.
If you freelance (design, illustration, photography, writing, or solo dev): which part of that would you most like to never think about again, and what did you try that didn't stick? Genuinely curious rather than pitching anything — I don't have a product, I'm trying to figure out whether the real pain is the bookkeeping itself or the decision-making around it (pricing, when to chase, what to set aside).
I work as an interpreter (independent contractor) and recently built an automation with google sheets and apps script with claude code, funny enough Claude end up being better than anything google offers for their platform.
Basically the agency I work pays every two weeks and the time between I receive an assignment and the the time i get paid for that assignment is about 45minutes.
I built an automation, that when I get a new assignment it automatically adds the assignment to my calendar, to the spreadsheet and it calculates the mileage using google maps API.
Then the spreadsheet calculates how much I’m gonna get in the following 2 paychecks and the one that’s currently running.
This mitigates problems with LLMs' low quality code. For a single user, the bar is much lower. The app doesn't have to have all the features that work for everyone's edge cases, implemented in a way that other devs can maintain. It only needs to be good enough for one user, and can be discarded and recreated when it gets too messy.
Programmers want you to review and own the code or they'll dismiss it as slop. Then the lawyers come in and say that doesn't give you copyright over the result. Damned if you do, damned if you don't.
Maybe it's better to just stop publishing stuff. If everything is uncopyrightable slop, the rational move is to just keep it to ourselves. So much for the free and open source software commons.
In my line of work, we convert words to numbers, and we have a codex that defines meanings for the numbers; that way, we can write publicly but (in making sure we choose words that, when converted to numbers, convert to numbers that have the second meaning we want) our public writing will only be understood on this second level with this second meaning only by those who use this system and codex. It's something like a form of real/human language cryptography: write for the world to see but only those who know the key will see what we're also saying to them. This was traditionally very hard to do for anyone who didn't grow up in a family that did this naturally from when they were kids. And I vibe-coded tools to make it easy. It's been amazingly helpful and worth its weight in gold :)
Hahaha! I love the list! Lots of professions have methods to send messages that simultaneously speak to various different audiences (while the bigger audiences don't know there are also other audiences as well, who will derive other meanings from the same message). The world is an interesting place full of surprises - and even to smart people, the text or image you are seeing might have meanings invisible to you, since you're not the intended audience for the secondary meaning.
Funny he said "Good APIs matter more than ever". My software for one these past few weeks has been my mega archive project, making an APIs & UIs out of 50+ providers whether they like it or not. I'm calling it clump chart. I even got into my turbotax TaxML files.
About four years ago I started building an RSS reader that was supposed to use an algorithm. The problem is, I didn't code in years and wasn't familiar with the modern web stack. I had to drop it, although was still able to sell the domain I got for $500 (nice).
And then, with Codex and Claude, I finally built it this year and enjoy it every day, along with 30 other strangers, since I made it public. It's a great feeling.
I now finally have an app that I use to save both links and articles to read, sort of in-between of Readwise Reader and Raindrop (and it does sync with Readwise).
I write my blog posts in a little browser-based note-taking app with sharing. It's minimal and looks great but has all I need.
I built a few Chrome extensions that I really wanted but couldn't find (well, couldn't find beautiful options).
I don't think everyone will do this. And getting the audience is a major pain, always has been, but now most of your early adopters just go and build their own things just like you. I'll still take it.
* And weather in Warsaw: https://github.com/stared/trmnl-x-weather-warsaw (here's location is hardcoded, so the expected use is fork and vive code to match both your location and visual preferences)
Many people build calorie / macro tracker like OP but they don't build maintenance calorie tracker.
Your app tracks calorie intake, doesn't keep track of total calorie burn of the day (maintenance calorie tracking)
And no, calorie burn from watch is not good enough!
>A 2020 systematic review by Fuller et al. found that consumer wearables often perform poorly at estimating energy expenditure, even when they measure steps and heart rate reasonably well. Across brands like Apple, Garmin, Polar, and Withings, calorie burn estimates frequently showed substantial error, often in the 15–40% range and sometimes exceeding 50%. Wearables can still be useful for tracking trends, but not for precise calorie calculations.
If the smartwatch data isn’t precise enough to estimate energy expenditure, what data does your website use that is so much better? Also, how much better it is? How do we even know it’s better? (I am not necessarily doubting it’s better, I am just interested in metrics you used)
The actual method used in the app is a bit more complex than this, but I think this explanation is good enough to convey the core idea for now. Later, I plan to publish a more detailed write up.
>Also, how much better it is? How do we even know it’s better?
Very simple! If a car salesman tells you that they use space technology in the engine to improve fuel efficiency, how do you know whether they are telling the truth?
You fill the tank with a specific amount of fuel and drive. If the fuel lasts much longer, you know the technology is actually improving efficiency.
The same idea applies to MacroCodex.
When you eat at maintenance you do not gain weight, when you eat above maintenance you gain weight, when you eat below it, you lose weight.
If you set a goal in the app to gain 1 lb per week and consistently eat the calorie target prescribed by the app, then your actual weight gain should match that rate. If it does, you know the maintenance calories calculated by the app are accurate.
You can choose to set "maintain" or "lose" or "gain".
Since the advent of these coding assistants, I gathered a list of software tool ideas I usually discarded due to lack of time. No more! Main one on that list: grocery assistant, with a barcode scanner for empty packaging to order new stuff, inventory management, coupling to recipes and most importantly: automatic ordering for delivery.
There's existing stuff for this but making a personal extension to this should be very well possible too.
Writing code with real use is very different from writing code for the sake of writing (looks cool, sounds cool).
Writing code with one user is the base case of a “why am I writing this” recursion - because I use it.
Every action leads to next insight.
For me, a browser extension that provides a table of content to chat interfaces because I always need to scroll up and down to read preview queries. Writing browser extension leads me to writing a RSS reader navigating with only keyboard. Writing RSS reader leads me to writing a virtual book shelf that visualize Goodreads record on a virtual 3D bookshelf. Writing virtual book shelf lead me to making a virtual climbing gym where I scan my local climbing gym into 3D model where I can map climbing videos on it. And it is just one chain
Especially when the primary motivation is not specifically monetization I find that I'm less likely to be discouraged when an action leads to an uncomfortable insight of "no one else would use this"
Building software steadily becomes accessible to non-developers. A freelance client I had now does most things alone with chatgpt. As a developer I have less and less the urge to build software. Instead I lately started doing more hardware stuff - flashing devices, setting up RPis and antennas, learning to solder. I hope I can keep my software dev job until retirement, but I‘m a bit pessimistic on that matter.
They basically inserted themselves between clients and developers. A few corporations ingested developer work and offered clients to pay a fraction, not giving anything back to developers.
Indeed, but it probably won‘t get any easier for us devs when clients also get easier access to open weight LLMs. I mean from personal experience 40 tok/s on an M3 pro with gpt-oss-20b holds up quite well for lots of tasks. Thinks are changing so fast.
Can confirm. I spent maybe 5-6 hours rebuilding an app I did in WinForms (originally Java… I forget the framework, Swing or something?) as a web app for a small business. Nothing fancy, basic CRM with existing database ported to SQLite and hosted on a cheapo VPS. Pure Codex CLI play none of that Cursor Orca rubbish. They’re very happy with it. Responsive, dark mode, works even without JavaScript. Nice unit test suite, simple server-side rendered. The traditional way. It’s the cats ass. Additional 2 hours for a couple of new requested features. Whatever! Easy peasy.
At work the team is leaning be heavily into internal tooling. Doesn’t need heavy support. Doesn’t need a fancy designer. Duct tape glue and liquid wrench are more than enough. It’s quick enough that it’s throw away and so if need be we can always write a v2 without losing months or years of work. Just enough to be useful, and then back to actual work.
There’s software for one and then there’s “Apple Watch type shit” which is tech that is highly personal and seems useful on paper but is absurdly extraneous - like calorie tracking, step counting, and being able to check things that are already easy to check.
Not for me.
I like software for people, and in terms of affordances and utility mass buy-in and standardization is part of the point.
I think it's sad that an article that to me was about sacrificing your own time to bring your family together has turned into sacrificing nothing to hyper optimise your life.
I'm building a podcast app just for me, because the podcast app I typically use (PocketCasts) is really good but there's just a couple of features I'd like and they don't bring out a lot of new features since it's not a monthly SaaS.
LLMs are a good help here because side projects typically slow down for me when I run into bugs / troubleshooting. So while I genuinely believe struggle is necessary for learning, being able to get over that hump to build something I need makes a big difference.
I'm not vibe coding the app though, since I actually want to know what it does. But I can see it being helpful for people who just need something and don't care about how it works.
> ...they don't bring out a lot of new features since it's not a monthly SaaS.
But it is! The Patron level of Pocket Casts is $15.99 / month!
I used to love Pocket Casts, but it didn't have the ability to easily add an RSS feed URL locally without submitting it to their index - which I don't want to do with my AI's personalised morning podcast (and that's another category, Podcast For One). I've had to use AntennaPod for that, but it's missing all the silence clipping and audio normalising features that Pocket Casts has. I'd love to see someone improve on it, and I've been tempted to do it myself too. Good luck!
This is such a romantic idea, but unfortunately Apple and Google are hostile to this model.
Of the two, iOS is /worst/ for running vibe coded apps.
You need to pay $99 a year to be an Apple developer if you want your app to live on your phone for more than a a week.
Distribution is a also pain in the ass even for personal use alone (you can't just toss around an IPA).
Finally, so many useful APIs are totally locked up. The fact that I cannot in 2026 extend Siri, for example, us infuriating.
Android is far more promising but Google recently shot themselves in the foot with their verified APK nonsense. Google's privacy posture is infamously nonexistent as well, and I have little interest in rooting a Pixel at my age.
Hopefully now that enough people can vibe code, these walls will crumble a bit.
I ran across WebXDC this past weekend, and I'm intrigued. I've built a few apps on it already, and I'm very pleasantly surprised by how cleverly inverted the whole model is (in combination with ChatMail). Delta Chat is available on iOS, I believe. I use Arcanechat on Android.
The beauty of the ecosystem is that apps are just zips of a webapp that is self-contained, and uses the webxdc api to message other copies of itself running on other folks' phones that are part of that chat group. Basically, the app communicates (and is distributed over!) the group chat (it's otherwise insulated from the internet). And because apps are just zipped webapps, they can be pulled apart, hacked on, and reassembled as needed for your group.
ChatMail relays tend not to retain info - they aren't hosts, they just relay encrypted messages (the platform mandates encryption). This sidesteps all the approval/review processes, and there is no app store or developer fee or even SDK. It's just folks trading apps and running them together.
It can't solve a lot of cases, but it solves some use cases (like a shared grocery list, or real-time card games, or games with leaderboards) really well.
You can make perfectly good apps that run great with a free account. They expire in a week instead of 3 months (for TestFlight builds). That's it. You can redeploy with minimal effort. You have to reauth with Apple every ~30 days. The reauth is the only part of that you can't have a cron job run (you need to authenticate the keychain).
For normie folks, the hoops here are mildly annoying. But if you're someone making vibe-coded apps, we're talking about two scripts and a cron job, here. Or just run Sideloadly, which I hear works pretty well.
There exist reasons to be annoyed by Apple's iOS security posture, but "I can't run my vibecoded apps" isn't an especially good one.
Indeed No thank you. Tried this with my personal habit tracker. Had to login again with my Apple ID so often and right when I needed my app it expired (good luck if you don’t have your laptop with you). Converted it into a PWA and be done with it. Weekly redeploys are a big PIA.
A pity you cannot use an iPhone for personal projects.
Which is why computer freedom is so important it needs to be enshrined into law. Stuff like remote attestation, locked down boot loaders and "app stores" need to be made literally illegal. Ownership of the computer should be a constitutional right.
A PWA might be well suited to your needs. Even Apple's restrictive model leaves plenty of room to build useful apps especially if they are designed to connect to a backend (like the ones in the post).
I totally hear you on that, but for my one off apps (and even some of my not one off apps) if I can't implement a feature in a PWA, I just learn to live without that feature.
App Store will be the place to download apps you can't avoid, like your bank's client.
For personal apps, the locked down stores are making themselves irrelevant. When you have software for one person, you don't want to publish it there. You want to bypass the gatekeepers.
It still holds for the desktop, though. LLMs have helped me quickly code new software for lab equipment (total users ~6), a research tool for my partner, and for several desktop apps I just use for me.
Two apps fill niches that I've wanted to code myself (one, I've wanted to code for the last twenty years on and off), but have never had the time to do so. Now, the speed from thought to working prototype is insane (less than an hour); it then takes a few days of occasional work to turn that prototype into something architecturally sound and half-decent.
And after that... well, I just keep improving, refactoring, refining, because it's so much fun.
For most people I believe it's a good exchange for the value of a more secure device.
But yes, I don't really want to pay Apple $100 a year just to run my apps. So I just have them as PWAs. Wouldn't work for many use cases, like when you need Apple Fitness data, but works for me and it looks fine.
Local first is a model I'm exploring. Combined with PWA it's pretty good and doesn't need to be native. LLMs are most useful for apps that are Good Enough™ anyway, and struggle with the remaining 5-10%.
Oh no! $99! If you're rich enough to have bought into the Apple walled garden, $99 is nothing. Can't go out for a nice dinner with the family for less than that these days. Yeah it would be great if Apple gave it away for free, but it's their walled garden that we've locked ourselves into. My bigger complaint about Apple is how overboard they've gone with the security stuff, making it it that much harder to develop software for their systems, on top of costing more. But, gotta go to where the users are.
Not a matter of money. It's an outright rejection of the "Billionaire-in-the-middle" model that has been eroding the traditional personal computing ownership model.
> No complaints over paying more than >$100 a month in subscriptions with coding agents
Plenty. The problem isn't so much the price as it is the loss of sovereignty.
It's not your model running on your computer, it's $CORPORATION's model running on their own data centers. They're just generously allowing you to use it, and only on their terms.
the post on BoopSnoop was very impactful for me too! As a software engineer, it's drilled into your brain that you have to think about high scalability and user facing products. I used to have a lot of ideas, but the first instinct was always to build a paid user facing product. None ever got built. But after reading that post a few years ago, my mindset was completely flipped. Since then I've built umpteen tools, services, macOS menubar apps, tail scale services .. just for myself. And I love it!
Genuine question: how will this "software for one" outlive its creator? I realize that if it were truly intended for just one user, it wouldn't matter, but in this specific case, the software is used by family members, so...
Yeah, this is a fear I keep bumping up against. Even relatively simple things like setting up pihole for my home network has pretty severe bus factor - if I get hit by a bus, no one else in the family would be able to fix the internet, short of chucking all the current hardware and asking the ISP tech to set up from scratch...
Even in early 2023 people were talking about this as a use case for ChatGPT copy-paste coding. Arvind Naranyan (AI As Normal Technology) used it to build edutainment applets for his kids.
Yup! Can confirm, I wanted to blog, but I wanted it to work in a certain way to let me focusing on writing that no single system had, or cost too much.
Now I'm actually doing it.
I've had an idea for how I want to be social with friends and family (a website, but setup within my vision). Now it's done
Could I have done all of these things on my own or just paid for existing services? Yes
But now AI can do it for me while I watch TV. I just show up to test.
I get it but just a word of advice - also look at industry standards like how SEO works as well. Otherwise your blog or content wouldn't even be discoverable.
But isn't it like super cheap to host a website in 2026? A lot of those builders are actual better if you just want to write. Otherwise you will have to maintain your website and also focus on writing, which would not allow you to just focus on writing.
I am not saying this out of my behind. I have done the exact same thing and ran into issues. If you just want to focus on content, building and maintain your own website can be a big distraction
Agree with the general spirit of the piece, but the examples that he gave create more problem than solving problems.
> A fitness app that sizes my smoothie to that morning's run
I don't feel I need more than the phone's calculator app for this particular problem.
Other examples could be solved by the Notes app, I feel.
I deliberately avoid creating new "apps", and make do with existing apps and solutions and create workflows rather than apps.
I write scripts, add features to scripts and understand all lines of them.
There are very few exceptions to this, like, I vibe coded a visual PDF cropper app that also handles scaling, merging in a GUI built with PyQt and GTK.
Having entire apps create maintenance burden and headaches.
To me it just sounds like we need an even higher level language for these sort of 'software for one' apps. There is still tons of boilerplate. Declarative is king.
I think that also extends to "software for few". It lowered the bar for building calculators and widgets for my users. Instead of lengthy explanations, I can just build little calculators and decision trees for my small user base.
Maggie Appleton said that it would enable "barefoot doctors" to bring helpful software to small communities. I have seen a lot of those underserved groups in my dev career.
I am not too happy about AI killing my business and inserting slop into every aspect of my life, but I am very hopeful about that aspect.
---
I also made some apps for one in the pre-AI days, and now I can quickly add features I've been procrastinating for ages. I can finally process data that I've been collecting but had no time to write code about.
It is never really for "One". If it is open source and published, It may be use by others... not to mention coding AI ML (Machine Learing), then for many.
I have written so much software for one... I cannot recall everything. I use some of it daily:
- I have my custom elf(glibc)/linux distro
- I use my own ffmpeg based x11 media player (will have an update for wayland).
- Since the "geniuses" at gogol blocked all accounts not using whatwg cartel web engines and blocked all self-hosted SMTP servers (I do not pay the DNS mob, then IP literals, which is stronger than SPF) to exchange with gmail.com users (now prisoners), I have my very own simple noscript/basic HTML servers (messaging, file transfer, openstreetmap browsing, etc). Unfortunately, most are C written from linux instead of assembly (at least, I removed most of the time the libc dependency with direct syscall programming). IPv6 is making everything internet software muuuuuuch easier to implement: I am lurking at an IPv6 only, super idiotic and real time voice/video/text/file transfer protocol, probably based on SIP (the main issue are randomly generated IPv6 addresses from mobile internet, which may require a mini-server for IPv6 address exchange, BAAAAAAAD!).
- Ofc, I still have my own minimal SMTP server (and a SMTP client to send emails), because there are still honnest people on internet. mutt forever.
- I am trying to build a RCS for linux, aka a Reduced Command Set with near direct hardwiring to linux syscalls. I lose comfort, but knowing that I removed giga tons of bloat feels so much satisfying it is worth it by light years. I will have to code my own command shell one day.
I still have C code since I currently need portability on IP-LOCKED ISAs (arm and x86-64). But I wish for RISC-V to be a success, namely as a modern, very often hitting the 'sweet spot' in technical design compromises, NON-IP-LOCKED ISA. I am writting much software in RISC-V assembly now... which I run on x86-64 with a small interpreter (written in x86-64 assembly)... trying to fix the bloat and kludge from corpo-like open source software (including the SDK).
Right now, I am writting my own wayland compositor using that framework. I was not expecting window management to be a bazillion of little things to do everywhere. And even if a wayland compositor is several orders of magnitude smaller and leaner than a x11 server, it ain't that a small project (ofc, I implemented myself the wayland wire protocol, no external libs here).
In my 'trying to fix the bloat and kludge' from corpo-like open source, I force myself to use my own exe/dynamic lib format for modern hardware (so lean, a simple RFC will be enough, and no more loader bloat/kludge). But I am scared at the abominations which are all the requirements of classic computer languages to run: I would like a mesa vulkan driver, but there is still c++ (super BAD) in there which makes porting to an alternative dynamic lib format a living hell, non trivial C won't be that easy neither (for instance the abomination of the ISO __thread keyword, amazing bright minds there). Taking perspective from that, I wonder why anybody sane would add on top of that a big runtime requirement??
And what people doing that have to keep in mind: IRL can put a violent stop to all of this.
> If you want fixes/features, you'll prompt your LLM and it'll make it so.
I've been using the agents since the day they came out. This, does not work. It works if you want an app that looks like it works if that works without any testing. If you use the app in any way that isn't boiler plate it will break. Because you can't prompt ai with a simple feature no matter how small and have it cover all the important cases.
What these models do is the absolute bear minimum to satisfy the user even if it means cheating. And anyone who doesn't look thinks its miraculous. It really is not.
So, you absolutely can prompt it to add a feature, but you need to prompt it 10-20 times to get the actual feature built. If you don't know this yet rip.
And before we continue, also known that when you prompt for the next feature. There is a high chance the last feature you built will be altered or deleted.
I'm not a sophisticated LLM user, but nor do I have tricky usecases.
Opus 4.5 with a plan mode is enough for me: it asks clarifying questions to map out the design space, it covers fiddly parts, it generates good unit/integration tests.
I have seen it cheat by deleting existing tests, since it's single-mindedly focused on the problem I give it. I've not used an LLM-to-review-an-LLM, but when I've used an LLM to review my own work, it notices when there are unrelated changes in my commit/working copy.
Besides, all of your critique is how I've heard product managers speak of human developers! And a tool needn't be perfect for it to be useful.
reply