20,193 karma · joined April 23, 2011
I'm a programmer and founder with experience ranging from web development to embedded systems and network programming.
Working as a principal engineer for a unicorn supply chain startup.
https://github.com/sarchertech/
I’ve never met you but based on your last 10 HN comments I think the world would be worse off without you in it.
Commercial speech doesn’t enjoy absolute protection.
They’re not far less common in average. Rates have gone down a bit with the annual flu shot, but there’s a few things going on.
-Flu rates vary wildly by year and decade. There were a few big pandemics where a very large percentage of people got the flu.
-There was no way to easily test for the flu, so people said they had it more frequently than they did.
In a given year about 12% of kids will come down with symptomatic influence which is not too far off from what we can estimate that historic numbers were in an average year.
>Strep
I looked up studies on this. Nothing conclusive, but likely no large change.
Recommendations about when to test have changed though because it turns out that if you culture every kid with a runny nose a large percentage of them will come back positive even though their symptoms are viral.
So you’re probably right that you notice fewer diagnosed strep cases, but there aren’t actually fewer symptomatic cases.
That’s very likely true.
> one big difference is indoor air ventilation standards improved dramatically, leading to less spread of disease indoors.
I don’t know if this is. In 1975 air conditioning was much less prevalent, so people were much more likely to open windows. Buildings were also much leakier.
And it’s not simple as building ventilation has steadily improved, so teasing out the impact is almost impossible. In the 80s there was a big push for energy efficiency, so we went from 15 cfm/person in 1973 to 5 in 1981 then back to 20 in 1989.
But building envelopes also got much tighter, so there’s less uncontrolled ventilation.
That’s just anyone who has small kids. Or works in healthcare, childcare, or any service industry really. It has nothing to do with 1975 vs 2025 (unless you’re one of the small amount of people who work remotely, and you don’t have kids or a spouse who falls into one of the above categories).
Also after a few years you stop getting so many colds.
So the vast majority of people have more space for entertaining than in 1975.
2. More remote communication
3. Fewer friends through church
4. Fewer friends through kid’s school
5. Kids have more homework and more structure activity than they used to.
6. Less need for entertainment
If that happens on a large enough scale those investments won’t be worth much.
Many studies find a small net drop in rent (1-5%). But other studies show no aggregate drop in rent. And importantly show an increase in rent for lower quality apartments.
Also for single family homes, nearby property values generally rise.
https://www.tandfonline.com/doi/full/10.1080/07352166.2026.2...
Plus other reasons like people want to live in good school districts and higher income areas tend to have better schools.
That ignores all the people who have other incentives to build.
Depending on the area you probably have about 10% of the people who have jobs that directly benefit from new construction (construction workers, plumbers, electricians etc…) or who are married to people who do. Then you have all the people who work or own businesses in sectors that indirectly benefit from new construction or directly benefit from increased density.
In my town of 8k people, I’ve seen 30 people from those groups show up in support of a new sub division vs half a dozen against.
> I don't even know how "vibe-coding" works in a team environment, because for me now it feels like it's "ideas to app" directly, so dumping by brain/ideas directly into a functional product.
If you have a system that does “ideas to app directly”, that certainly sounds like what people are talking about when they say software factory. In fact my very large company would probably pay you millions if you could come in and make this work for our applications.
I have a practically unlimited AI budget and I use it all day long. We have entire teams of very smart people trying to make it possible for PMs to turn jira tickets into features.
We ain’t there yet.
My strong suspicion is that we won’t get there without something close to AGI. And if we get there, there won’t be any white collar jobs left much less software developer jobs.
If AI can take your raw idea and turn it into code with no input from you, it will be able to generate the ideas too.
> I agree with your linked post, and that is sort-of a big issue
This was last week with frontier models. Practically what it means is that I need to audit the code. Approximate changes don’t work when you move past very small scale projects.
That’s the exact opposite of the experience I’ve had in my town.
It seems that money is the primary driver of turnout at town council meetings. I live at the top of a mountain with 1 real road and 2 smaller frequently closed roads down to the city below. Developers really want to build up here, but there’s no way to widen the road, and no realistic way to build more. There’s not enough flat land to build enough jobs, shopping, dining etc… up here, so increasing density just means more cars making the trip 30 minutes into the city.
Every single time there’s an issue about development, you get dozens of real estate agents, developers, and people who work in construction showing up because their bank account depends on it. Every local election, you have candidates bankrolled by out of town developers.
“No” is much less well organized and will only occasionally turn out in larger numbers than “Yes” despite dominating in elections.
>It took probably 1 month to implement what would have taken 1 year without AI.
1. I’m not saying that AI can’t speed you up. I’m only saying that you can’t ignore the code for anything beyond a toy app (at least not sustainably).
2. How much is that is down to motivation though? You’ve been developing something for 15 then years then suddenly there’s a brand new development methodology that is fun to mess around with and addictive.
I wrote about some of the issues I uncovered auditing my own vibe coded project in another thread https://news.ycombinator.com/item?id=49856310
If I had a working software factory like you describe, I’d expect you to have hundreds of apps of that level of complexity in a year.
> I don't even know how "vibe-coding" works in a team environment, because for me now it feels like it's "ideas to app" directly, so dumping by brain/ideas directly into a functional product.
But you don’t have any users much less paying users, so you have no idea if this system works when you do.
Now come back to me when you have paying users.
Better yet come back when you have paying users who depend on your app to do their job. And in addition to buttons changing location, you are constantly breaking their work flows because they are using the app in ways you didn’t anticipate.
Overly defensive code is harder to read and change for both humans and LLMs.
And many times it makes debugging harder by moving or suppressing failures.
Come back to me when your app has users and adding new features subtly (or not so subtly) breaks every work flow that you haven’t explicitly tested.
You can’t commit the prompt and regenerate the source code each time because the whole reason that an LLM is useful is that it makes thousands of decisions for you. And those decisions are different each time you regenerate.
The only way to enforce that those decisions are the same each time you regenerate is to encode all of them in tests. But at even moderate complexity that leads to an overconstraint problem that will halt development.
We see this when using LLMs on large apps. Anthropic gave up on their C compiler. Even with an unlimited budget they stopped being able to make progress on it.
I see this in some games I made for my 4 year old. I had them one shot some “juice” when he gets an addition problem right. Combination of screen shake, sounds, flashing light, explosions etc…
It looks pretty cool, but when I tried to tweak the animations with prompts it was always worse. I eventually went in and edited the code myself and I could see why it was so hard for the LLM to change anything because it was a horrific mess of interwoven animations.
I was able to pull everything apart and manually adjust what I wanted.
I mostly vibe coded a queuing system to replace something we’re using at work (last week. Spent about $1500). Then I meticulously went through the code.
It was much harder to review because it was ultra defensive and included guards for tons of edge cases that weren’t possible.
Unnecessary abstractions for possible extension later. Useless indirection. Probably 3x as much code as there would have been if I’d written it by hand.
I didn’t one shot this. I kept a pretty tight leash on the AI. I had probably a dozen markdown files with of plans that I created over hours of back and forth with the AI and reviewed before each implementation round. I had automated reviews and quality gates etc…
What I found in review was that it was full of very subtle bugs that would have bitten hard in prod. Committing offsets asynchronously that would lead to dropped messages. Clock drift bugs that would lead to dropped messages or write amplification storms. Lack of back pressure in some stages of the pipeline that would cause notes to get silently OOM killed. Weird over-insistence on never crashing in most places that would mask systemic errors.
If I’d just shipped it without review, it would have mostly worked. But at the scale it’s going to be used (tens of thousands of messages per second) it would have caused production issues for months while we tracked down each of these issues.
> why I would review 100k's of lines of code.
If the core or that app is more than a few thousand lines of code, something is seriously wrong.
I don’t want to shit on your app. It’s cool. I’m glad you built it. I’ve vibe coded all kinds of toy apps for myself and my kids.
But it’s not strong evidence that code is irrelevant.
1. If what you say is working, you have a working software factory that should be capable of matching the output of dozens of engineers.
What very impressive externally verifiable results have you had with this?
2. If you are working in software with plenty of customers, my strong suspicion is that there are people on your team who are looking at the code who furiously trying to reign in your output.
>barely even tested
So the lowest stakes possible and you have absolutely no idea what bugs are waiting.
I think it’s more likely you’re just exaggerating. How many is many? If you personally know more than 5 doctors with $500k-$750k in debt after med school I’ll eat my hat.
That reddit link only had 7 actual doctors who claimed to have graduated with more than $500k in debt. In a post about high debt with 200+ comments in a subreddit for doctors.
And the OP mentioned in the comments that he was borrowing enough to pay for his “expensive” med school and to support himself, his stay at home wife, and their baby in an expensive area.
> Residency match rates for American graduates from American MD or DO programs are more like 93%
That’s only after primary march. Match rate after SOAP is 98%. Roughly half of the unmatched 2% get marched in the next max cycle which is where I got 99%. Of the remaining 1% most of those want to eventually place somewhere.
>For American IMGs
Yeah don’t go to a factory medical school in the Caribbean. It’s a bad idea.
It’s definitely almost 0 financial risk. The only real risk is that you just hate being a a doctor. But even in that case med school unlocks tons of opportunities outside of clinical practice.
The chance of you ending up suffering financial hardship is too small to consider in your decision making.
The real out of your control risks would be severe disability that prevents you from finishing med school or working (before you can afford disability insurance). But in the case it really doesn’t matter whether you owe $50k from undergrad or $250k from med school. You’re not paying either one back and you can actually bankrupt out of student loan debt in those cases.
What you’ve done is constructed an argument where no matter what position the AMA takes, they are the bad guys.
> salaries depended on my lobbying taking its time to produce as little of a result as possible
This doesn’t make sense. How would the AMA board communicate this to their voting members when all of their communications say that they doing all they can to increase resident funding.
How about the first board to vote to change their position 30 years ago? How would the voting members know that they were planning on slow walking their new position?
Do you think they have a secret doctor only communication channel where they tell them “don’t worry guys we’re not trying that hard”. Because without that if your premise is correct all of the members would have voted them out ASAP.
Also how is the AMA (which only represents about 15% of doctors btw) preventing hospitals from effectively lobbying for an increase? They have a huge financial interest in more resident funding.