It is incredibly easy now to get an idea to the prototype stage, but making it production-ready still needs boring old software engineering skills. I know countless people who followed the "I'll vibe code my own business" trend, and a few of them did get pretty far, but ultimately not a single one actually launched. Anyone who has been doing this professionally will tell you that the "last step" is what takes the majority of time and effort.

Does the data not support a 2X increase in packages?

Pre-ChatGPT, in ~2020, there were about 5,000 new packages per month. Starting in 2025 (the actual year agents took off), there is a clear uptick in packages that is consistently about 10,000 or 2X the pre-ChatGPT era.

In general, the rate of increase is on a clear exponential. So while we might not see a step change in productivity, there comes a point where the average developer is in fact 10X productive than before. It just doesn't feel so crazy because it can about in discrete 5% boosts.

I also disagree with the dataset being a good indicator of productivity. I wouldn't actually suspect the number of packages or the frequency of updates to track closely with productivity. My first order guess would that AI would actually be deflationary. Why spend the time to open source something that AI can gen up for anyone on a case by case basis specific to the project. it takes a certain level of dedication and passion for a person to open source a project and if the AI just made it for them, then they haven't actually made the investment of their time and effort to make them feel justified in publishing the package.

The metrics I would expect to go up are actually the size of codebases, the number of forks of projects that create hyper customized versions of tools and libraries, and other metrics like that.

Overall, I'd predict AI is deflationary on the number of products that exist. If AI removes the friction involved with just making a custom solution, then the amount of demand for middleman software should actually fall as products vertically integrate and reduce dependencies.

Maybe the top 15,000 PyPi packages isn't the best way to measure this?

Apparently new iOS app submissions jumped by 24% last year:

> According to Appfigures Explorer, Apple's App Store saw 557K new app submissions in 2025, a whopping 24% increase from 2024, and the first meaningful increase since 2016's all-time high of 1M apps.

The chart shows stagnant new iOS app submissions until AI.

Here's a month by month bar chart from 2019 to Feb 2026: https://www.statista.com/statistics/1020964/apple-app-store-...

I deleted vscode and replaced with a hyper personal dashboard that combines information from everywhere.

I have a news feed, work tab for managing issues/PRs, markdown editor with folders, calendar, AI powered buttons all over the place (I click a button, it does something interesting with Claude code I can't do programmatically).

Why don't I share it? Because it's highly personal, others would find it doesn't fit their own workflow.

Claude Code was released for general use in May 2025. It's only March.

Also using PyPI as a benchmark is incredibly myopic. Github's 2025 Octoverse[0] is more informative. In that report, you can see a clear inflection point in total users[1] and total open source contributions[2].

[0]: https://github.blog/news-insights/octoverse/octoverse-a-new-...

[1]: https://github.blog/wp-content/uploads/2025/10/octoverse-202...

[2]: https://github.blog/wp-content/uploads/2025/10/octoverse-202...

AI makes the first 90% of writing an app super easy and the last 10% way harder because you have all the subtle issues of a big codebase but none of the familiarity. Most people give up there.

Not sure that I'd look at python package stats to build this particular argument on.

First, I find that I'm using a lot fewer libraries in general because I am less constrained by the mental models imposed by library authors upon what I'm actually trying to do. Libraries are often heavy and by nature abstract low-level calls from API. These days, I'm far more likely to have 2-3 functions that make those low-level calls directly without any conceptual baggage.

Second, I am generalizing but a reasonable assertion can be made that publishing a package is implicitly launching an open source project, however small in scope or audience. Running OSS projects is a) extremely demanding b) a lot of pain for questionable reward. When you put something into the universe you're taking a non-zero amount of responsibility for it, even just reputationally. Maintainers burn out all of the time, and not everyone is signed up for that. I don't think there's going to be anything remotely like a 1:1 Venn for LLM use and package publishing.

I would counter-argue that in most cases, there might already be too many libraries for everything under the sun. Consolidation around the libraries that are genuinely amazing is not a terrible thing.

Third, one of the most recurring sentiments in these sorts of threads is that people are finally able to work through the long lists of ideas they had but would have never otherwise gotten around to. Some of those ideas might have legs as a product or OSS project, but a lot of them are going to be thought experiments or solve problems for the person writing them, and IMO that's a W not an L.

Fourth, once most devs are past the "vibe" party trick phase of LLM adoption, they are less likely to squat out entire projects and far, far more likely to return to doing all of the things that they were doing before; just doing them faster and with less typing up-front.

In other words, don't think project-level. Successful LLM use cases are commit-level.

Theres tons of ai apps. They're all general use chatbots or coding agents. Manus, Cursor, ChatGPT. Almost every app that has a robust search uses a reranker llm. AI is everywhere.

As far as totally new products - I built one (Habit.am - wordless journaling for mental health) and new products require new habits, people trying new things, its not that easy to change people's behavior. It would be much easier for me to sell my little app if it was a literal plain old journal.

AI does make me more productive. At least until the stage of getting my idea to the "working prototype stage". But in my personal experience, no one has been realistically able to get to the 10x level that a lot of people claim to have achieved with LLMs.

Yes, you do produce more code. But LoC produced is never a healthy metric. Reviewing the LLM generated code, polishing the result and getting it to production-level quality still very much requires a human-in-the-loop with dedicated time and effort.

On the other hand, people who vibe code and claims to be 10x productive, who produces numerous PRs with large diffs usually bog down the overall productivity of teams by requiring tenuous code reviews.

Some of us are forced to fast-track this review process so as to not slow down these "star developers" which leads to the slow erosion in overall code quality which in my opinion would more than offset the productivity gains from using the AI tools in the first place.

I fail to see why the author thinks Python packages are a good proxy for AI driven/built code. I've built a number of projects with AI, but I haven't created any new packages.

It's like looking at tire sales to wonder about where the EV cars are.

This is going to cause people to react, but I think those of us that truly love opensource don't push AI generated code upstream because we know it's just not ready for use beyond agentic use. It's just not robust for alot of use common use cases because the code produces things that are hyper hardcoded by default, and the bugs are so basic, i doubt any developer that actually cared would push something so shamefully sloppy upstream with their name on it.

The tools for generating AI code aren't yet capable of producing code that is decent enough for general purpose use cases, with good robust tests, and clean and quality.

The thesis has it backwards. We will see fewer published/downloaded apps/packages as people rely on others less. I'm not sure we're quite there yet but I'm increasingly likely to spend a few minutes giving an LLM a chance to make a tool I need instead of sifting through sketchy and dodgy websites for some slightly obscure functionality. I use fewer ad-heavy sites that for converting a one text file format to another.

Personally, I see the paid or adware software market shrinking, not growing, as a testament to the success of LLMs in coding.

This remains me so much of the .COM bubble in 2000. A lot of clueless companies thought that they just need to “do internet” without any further understanding or strategy. They burned a ton of money and got nothing out of it. Other companies understood that the internet is an enabling technology that can support a lot of business processes. So they quietly improved their business with the help of the internet.

I see the same with AI. Some companies will use AI quietly and productively without much fuzz. Others are just using it as a marketing tool or an ego trip by execs but no real understanding.

I won't make any claims as to the Python ecosystem and why there is no effect seen here (and I suppose no effect seen of the Internet on productivity) but one thing that is entirely normal for me now is that I never see the need to open-source anything. I also don't use many new open-source projects. I can usually command Claude Code to build a highly idiosyncratic thing of greater utility. The README.md is a good source of feature inspiration but there are many packages I simply don't bother using any more.

Besides, it's working for me. If it isn't working for others I don't want to convince them of anything. I do want to hear from other people for whom it's working, though, so I'm happy to share when things work for me.

Where are they? Well they aren't being uploaded to PyPI. 90% of the "AI apps" one-off scripts that get used by exactly one person and thrown away. The rest are too proprietary, too personal, or too weird to share.

I’ve done a event ticket system that’s in production. Stripe integration, resend for mailing and a scan app to scan tickets. It’s for my own club but it’s been working quite well. Took about 80 hours from inception to live with a focus on testing.

I’ve done some experiments with reading gedcom files, and I think I’m quite close to a demoable version of a genealogy app.

Biggest thing is a tool for remotely working musicians. It’s about 10000 lines of well written rust, it is a demoable state and I wish I could work more on it but I just started a new job.

But yeah, this wouldn’t have been possible if I hadn’t been a very experienced dev who knows how to get things live. Also I’ve found a way to work with LLMs that works for me, I can quickly steer the process in the right way and I understand the code thats written, again it’s possible that a lot of real experience is needed for this.

Vibe coding is actually a brilliant MLM scheme: people buy tokens to generate apps that re-sell tokens (99% of those apps are AI-something).

Looking at Python packages, or any developer-facing form of software, is not a good indicator of AI-based production. The key benefit of AI development is that our focus moves up a few layers of abstraction, allowing us to focus on real-world solutions. Instead of measuring Github, you need to measure feature releases, internal tools created, single-user applications built for a single niche use case.

Measuring python packages to indicate AI-based production is like measuring saw production to measure the effectiveness of the steam engine. You need to look at houses and communities being built, not the tools.

The first 80% is the easy part, and good ol' Visual Basic was fabulous at it, but the last 80% is the time suck.

Same with vibe-coded stuff.

The reason why the release cadence of apps about AI has increased presumably reflects the simple facts that

a) there are likely many more active, eager contributors all of a sudden, and

b) there's suddenly a huge amount of new papers published every week about algorithms and techniques that said contributors then eagerly implement (usually of dubious benefit).

More cynically, one might also hypothesize that

c) code quality has dropped, so more frequent releases are required to fix broken programs.

I've been vibe-coding a Plex music player app for MacOS and iOS. (I don't like PlexAmp) I've got to the point where they are the apps I use for listening to music. But they are really just in an alpha/beta state and I'm having a pretty hard time getting past that. The last few weeks have felt like I'm playing wack-a-mole with bugs and issues. It's definitely not at the point others will be willing to use it as their daily app. I'm having to decide now if I keep wanting to put time into it. The vibe-coding isn't as fun when you're just fixing bugs.

The bottleneck shifted but didn't disappear. Getting to a working prototype in a weekend is real, but error handling, edge cases, and ops work hasn't gotten much faster. Distribution is completely unchanged too. A lot of these 'where are the AI apps' questions are really asking why there aren't more successful AI businesses, which is a harder and very different problem.

By "apps" this author apparently means "PyPi packages". This is a bafflingly myopic perspective in a world of myopic perspectives. Do we really expect people vibecoding "apps" to put anything on PyPi as a result? They're consumers of packagers, not creators.

I don't blame people for responding to the title instead of the article, because the article itself doesn't bother to answer its own question.

I like using it to make personal apps that are specific to my use-case and solve problems I've had for ages, but I like my job (scientist), and I don't want to run an app company.

Wouldn't the apps go into the Apple store and Android play? I guess looking at python packages is valid, but I don't think it's the first thing someone thinks to target with vibe coding. And many apps go to be websites, a website never tells me much about how it is made as a user of the site.

I am learning music. I used codex to create a native metronome app, a circle of fifths app, a practice journal app. I try to build a native app alternatives.

I have no plans of publishing them or making the open source, so it will not be a part of this metric. I believe others are doing this too.

well, many apps i made are really good but i would never bother to share it, takes unnecessary effort and i don't really know what works best for me will work like that for others

maybe some developers are more productive while the rest of em is laid off.. keeping the same release cadence but with fewer devs?

i know maybe this is not to your analysis as its about open source stuff, but this is the sentiment i see with some companies. rather than have 10x output which their clients dont need, they produce things cheaper and earn more money from what they produce. (and later lose that revenue to a breach :p)

    > So, let’s ask again, why? Why is this jump concentrated in software about AI?...Money and hype

The AI field right now is drowning in hype and jumping from one fad to another.

Don't get me wrong: there are real productivity gains to be had, but the reality is that building small one-offs and personal tools is not the same thing as building, operationalizing, and maintaining a large system used by paying customers and performing critical business transactions.

A lot of devs are surrendering their critical thinking facilities to coding agents now. This is part of why the hype has to exist: to convince devs, teams, and leaders that they are "falling behind". Hand over more of your attention (and $$$) to the model providers, create the dependency, shut off your critical thinking, and the loop manifests itself.

The providers are no different from doctors pushing OxyContin in this sense; make teams dependent on the product. The more they use the product, the more they build a dependency. Junior and mid-career devs have their growth curves fully stunted and become entirely reliant on the LLM to even perform basic functions. Leaders believe the hype and lay off teams and replace them with agents, mistaking speed for velocity. The more slop a team codes with AI, the more they become reliant on AI to maintain the codebase because now no one understands it. What do you do now? Double down; more AI! Of course, the answer is an AI code reviewer!. Nothing that more tokens can't solve.

I work with a team that is heavily, heavily using AI and I'm building much of the supporting infrastructure to make this work. But what's clear is that while there are productivity gains to be had, a lot of it is also just hype to keep the $$$ flowing.

A friend of mine who is tech savvy and I would say has novice level coding experience decided to build his dream app. Its really been a disaster. The app is completely broken in many different ways, has functionality gaps, no security, no thought out infrastructure, its pretty much a dumpster fire. The problem is that he doesn't know what he doesn't know, so its impossible for him to actually fix it beyond instructing the AI over and over to simply "fix it". The more this is done, the worst the app becomes. He's tried all the major AI vendors, from scratch, same result, a complete mess of code. He's given up on it now and has moved on with his life.

Im not saying that AI is bad, infact, its the opposite, its one of the most important tools that I have seen introduced in my lifetime. Its like a calculator. Its not going to turn everyone into a mathematician, but it will turn those who have an understanding of math into faster mathematician.

On Show HN.

My guess - these are not not on PyPI because of libraries. AI generating is good when you don't care about how your app works, when implementation details does not matter.

When you are developing library it's exact opposite - you really care about how it works and which interface it provides so you end up writing it mostly by hand.

So far, as sideloaded APKs on my tablet. Most recently one that makes it easier to learn Dutch and quiz myself based on captions from tv shows

> So where are all the AI apps?

They're in the app stores. Apple's review times are skyrocketing at the moment due to the influx of new apps.

While a good post the title is a bit ambiguous. The post is about applications created using AI not applications with AI functionality embedded.

I’ll ask another question. Why isn’t software getting better? Seems like software is buggier than ever. Can’t we just have an LLM running in a loop fixing bugs? Apparently not. Is this the future? Just getting drowned in garbage software faster and faster?

I have a number of small apps and libraries I've prompted into existence and have never considered publishing. They work great for me, but are slop for someone else. All the cases I haven't used them for are likely incomplete or buggy or weird, the code quality is poor, and documentation is poor (worse than not existing in many cases.)

Plus you all have LLMs at home. I have my version that takes care of exactly my needs and you can have yours.

I absolutely hate web development with a passion and haven’t done a new from the ground up web app in 25 years and even since then it was mostly a quick copy and paste to add a feature.

But since late last year even when it’s not part of the requirements leading app dev + cloud consulting projects, I’ll throw in a feature complete internal web admin site to manage everything for a project with a UI that looks like something I would have done 25 years ago with a decent UX.

They are completely vibe coded, authenticated with Amazon Cognito and the only things I verify are that unauthenticated users can’t access endpoints, the permissions of the lambda hosting environment (IAM role) and the database user it’s using permissions.

Only at most 5 people will ever use the website at a time - but yeah I get scalability for free (not that it matters) because it’s hosted on Lambda. (yes with IAC)

The website would not exist at all if it weren’t for AI.

Now just to be clear, if a website is meant for real people and the customer’s customers. I’ll insist on a real web designer and a real web developer be assigned to the project with me.

Why would package be used as the standard? What person fully leveraging AI is going to put up packages for release? They (their AI model) write the code to leverage it themselves. There is no reason to take on the maintenance of a public package just because you have AI now. If anything, packages are a net drag on new AI productivity because then you'd have to worry about breaking changes, etc. As far as actual apps being built by AI, the same indie hackers that had garbage codebases that worked well enough for them to print money are just moving even faster. There are plenty of stories about that.

We’re in a personal software era. Or disposable software era however you want to look at it. I think most people are building for themselves and no longer needing to lean on community to get a lot of things done now.

Self plug, but basically that’s the TL;DR https://robertdelu.ca/2026/02/02/personal-software-era/

Hmmm, my anecdotal experience doesn't match up with this article. Personally I am seeing an explosion of AI-created apps. A number of different subreddits I use for disparate interests have been inundated with them lately. Show HN has experienced the same thing, no?

Cool data. What do I do with it? None of my use cases involve writing software, so I don't think this is _for_ me since my extensive AI use wouldn't show up in git commits, but I'm not sure who it's for. When I'm talking to artist friends, musician friends, academic friends, etc data is nice to have but I'm talking in stories: the real thing I did and how it made me better at the thing.

AI is unbelievably useful and will continue to make an impact but a few things:

- The 80/20 rule still applies. We’ve optimized the 20% of time part (a lot!) but all the hype is only including the 80% of work part. It looks amazing and is, but you can’t escape the reality of ~80% of the time is still needed on non-trivial projects.

- Breathless AI CEO hype because they need money. This stuff costs a lot. This has passed on to run of the mill CEOs that want to feel ahead of things and smart.

- You should be shipping faster in many cases. Lots of hype but there is real value especially in automating lots of communication and organization tasks.

Stuck behind Apple's app review process.

Intel AI denoiser in Blender.

I feel that your assumption that everyone will want to share is a flawed one.

I, for one, am not publishing my “apps” for others to use because my “apps” make me money

This article is very poorly researched and reasoned, but it's in the "AI hater" category so I guess it's no surprise it's on the front page.

Number of iOS apps has exploded since ChatGPT came out, according to Sensor Tower: https://i.imgur.com/TOlazzk.png

Furthermore, most productivity gains will be in private repos, either in a work setting or individuals' personal projects.

I agree with the premise of the article, in the sense that there has not been, and I don't think there will be, a 100x increase in "productivity".

However, PyPi is not really the best way to measure this as the amount of people who take time to wrap their code into a proper package, register into PyPi, push a package, etc... is quite low. Very narrow sampling window.

I do think AI will directly fuel the creation of a lot of personal apps that will not be published anywhere. AI lower the barrier of entry, as we all know, so now regular folks with a bit of technical knowledge can just build the app they want tailored to their needs. I think we´ll see a lot of that.

here: https://www.youtube.com/shorts/vGKC9LpGnOQ

On my local computer used only by me because now I don't need a corporation to make them for me. In the past decades I'd make maybe one or two full blown applications for myself per 10 years. In the past year "I" (read: a corporate AI and I) have made dozens to scratch many itches I've had for a very long time.

It's a great change for a human person. I'm not pretending I'm making something other people would buy nor do I want to. That's the point.

I am now scared to talk to anyone. Eventually the conversation turns to AI and they want to talk or show their vibecoded app.

I am just tired boss. I am not going to look at your app.

This is so stupid. I don't know whether AI has improved things but this is clearly cope, we're not even a year into the transition since agentic coding took over so any data you gather now is not the full story.

But people are desperate for data right? Desperate to prove that AI hasn't done shit.

Maybe. But this much is true. If AI keeps improving and if the trendline keeps going, we're not going to need data to prove something equivalent to the ground existing.

This is such copium for AI haters. I stopped working almost any single line of code at the beginning of this year and I've shipped 3 production projects that would have taken months or years to build by hand in a matter of days.

Except none of them are open source so they don't show up in this article's metrics.

But it's fine. Keep your head in the sand. It doesn't change the once in a lifetime shift we are currently experiencing.

[dead]

[dead]

[dead]

[dead]

[flagged]

No one needs another SaaS. Games are the real killer app for AI. Hear me out.

I've wanted to make video games forever. It's fun, and scratches an itch that no other kind of programming does. But making a game is a mountain of work that is almost completely unassailable for an individual in their free time. The sheer volume of assets to be created stops anything from ever being more than a silly little demo. Now, with Gemini 3.1, I can build an asset pipeline that generates an entire game's worth of graphics in minutes, and actually be able to build a game. And the assets are good. With the right prompting and pipeline, Gemini can now easily generate extremely high quality 2d assets with consistent art direction and perfect prompt adherence. It's not about asking AI to make a game for you, it's about enabling an individual to finally be able to realize their vision without having to resort to generic premade asset libraries.

All apps you are using are made with AI.

Not all of us get addicted to the rat race and wake up at 3am to run more Ralph loops. Some are perfectly content getting the same amount of work done as before, just with less investment of time and effort.