I use Netlify connected to Github/Gitlab, which does a build on every commit to the main branch. For low-volume static sites (which all of mine are) this is free, fast, and convenient.
I'm actually using Github issues as a low-volume/high-latency DB for a project right now. I use a Netlify function to do extremely simple auth and then open an issue on Github, which triggers a CI action, which appends to a .jsonl file in the repo (different branch), which triggers a Netlify build, which then serves the new jsonl file. (If the repo were public, this wouldn't even have to trigger a rebuild of the website, it could just get the .jsonl from Github directly).
It's totally hack-a-licious but it also totally works for my N<10 users.
ACA is not Universal Health Coverage. So this would be different enough. And you don't need perpetual majority control, only sensible politicians for about a generation (like FDR/Truman/Eisenhower/JFK) to lock in a popular entitlement.
FDR and "sensible" is always funny to me. The sensible politician that imprisoned Japanese Americans, destroying the lives of 120,000 of our fellow citizens.
The problem with this idea is that knowing Python makes the model a better Swift programmer, as does a higher-number of parameters during training. So you'd be so much better off with a 90B general purpose model trained on everything anyway.
Except assuming a fixed budget of parameters, there is clearly stuff that's better for programming than others. E.g. Qwen 27b is a better coding model than Gemma 4 31B. More params doesn't automatically win. Perhaps what makes a good swift programming model is a ton of python training, so those two things can't be separated - but that doesn't negate the idea of loading a model that's good at the specific task you want - or has specific knowledge of the libraries and tools for the language you're using at the expense of the ones you're not using.
I don't see how you could verify "same application in every meaningful way" inside of 30 minutes. Even if the other fellow just vibed it up like you did, him caring about the product and doing thoughtful QA and actually using the application and making little tweaks and staking his reputation on it and committing to its ongoing maintenance is what matters. You may even have an identical application at the end of the day but I'll pay $20 for one that I'm better able to trust.
> I don't see how you could verify "same application in every meaningful way" inside of 30 minutes.
That was not difficult. Though obviously I wouldn’t declare it so and actually ship it to paying people in 30 minutes.
As for the rest of your point, that’s certainly going to be a fair assessment, much of the time, for now.
But that wasn’t really the point. We’re already pretty close to being there on “just tell the machine what you want” for huge swaths of non-critical software. And I don’t think pressure from that is going to let up.
This is what I'm saying constantly to people, I recognize it makes me fantastically obnoxious, but also, I'm gesturing wildly to the things that are happening and yelling, "THINGS DON'T HAVE TO BE THIS WAY."
Then again, I'm also building my cabin in the woods.
Better example, leetcode interviews became so popular because people were memorizing FizzBuzz and they needed an alternative (but said interviewers didn't fully understand the purpose of FizzBuzz and though "harder means better, right?").
leetcode interviews be ame a thing, because many believed solving leetcode style puzzles makes you superior. Not because someone would memorize fizzbuzz.
This thinking was popularized by coding competions which existed before leetcode and inspired leetcode. Schools themselves produced people who believed these puzzles are what makes you superior developer - one of us, special and choosen.
You're right about those things, but you're forgetting that we had no source control, and almost all software was closed source. You took what you were given and you liked it, and you developed strong attachments to particular versions of software which you clung to well beyond its use-by date (like Python 2.7). MS-DOS 3.3 forever!
In the 80s, source control was like networking. Yes some orgs had it (mostly academia) but the rest of us just had a precious floppy that we copied the known-good source onto.
reply