A lot of the product misses come from large product teams where nobody owns anything, and where there's minimal internal goal alignment. I know that, at work, what I am incentivized to do has nothing to do with what I consider the most important work I could do.... and I also know that the maze I would have to cross to be allowed to do the better work is just not worth it.
The smaller the org, the better the chances of having only reasonable levels of misalignment... but without taking on high risk, it's just better pat to work on a large, worse org. Frustrating.
In large orgs, each team spends a ridiculous amount of effort adding the “glue” logic to work around design/architectural issues with the other teams’ products they interface with.
Neither one really understands the other’s component.
Cooperation is almost impossible since such would derail one project manager’s unrealistic schedule and demean their authority.
Just a little gripe, but near the end of the article he links to a site for something called "Omarchy" which appears to be a Linux distribution. If you click on the link, the only information available is in the form of a video.
If you're making a Linux distro, please don't do this. I will not watch your video. I was interested, and now I'm not.
Complexity + a lack of incentives to take time to get things right.
Imo we know the tools (social, + technical) to make good software that works. But in the face of massive complexity, people just want the next feature, the next thing sales can use as a win
And yeah I agree with the other commenter here - I don't think this is AI. We had terrible software before, even declining quality.
> As I’m writing this, we’re in the middle of an AI-induced mass psychosis.
I work in mediation and have recently seen a few cases where I’ve wondered if the participants were suffering from AI psychosis as related to their claims: where they’d constructed their legal cases hand-in-hand with an LLM and had become absolutely convinced despite a contestable set of facts.
People have always been subject to delusion, but in these cases, all documentation submitted was obviously created by chatbots.
I've also noticed so many pieces of software becoming worse and worse, either with useless features (basically any new thing on Slack in the past year) or with a huge amount of large bugs, over the past year.
It's glaringly obvious that poorly checked, AI-written, code is the culprit, I'm just wondering when companies will wake up.
Nobody gives a flying fuck about resiliency, good practices, security or any of that now that AI is there. All anyone cares about is pure output of features.
Any and all "productivity" gains attributed to AI efficiency are actually due to the complete disregard for any care or standard in the software industry in the past 2 years. Yeah you can output more shit when you don't give a fuck, big surprise, AI doesn't change anything here.
At the end of the day it has nothing to do with AI. People have been bitching about bugs in it for a long time. It's companies that give zero shits about releasing bugs.
This comes back to the users. As long as they keep buying the shit software why spend money on QA.
I mean, you're the problem. Did you stop using slack? Are they taking losses because of it? If not, why change?
Right now, the majority of companies I interact with seems to look at their now (let's assume) increased capabilities, and turns up the speed. But there's other knobs to turn up instead. Robustness, quality, user experience.
Optimistically, I think we might just enter a new phase at some point, where businesses make very conscious decisions about what to invest in to differentiate. Like, everyone has support bots now. Where do we invest, absolutely stunning human support? Or is support not that key and we should invest into being the most reliable? The most enjoyable to use?
Those two twitter links with people that got hospitalized might just be people trying to get their 5 seconds of internet fame by (pretending to be) doing something dumb.
So taking it at face value is kinda part of the problem here.
The only way this can get better is by completely suffocating it off attention.
It doesn't exist. It's not real.
I agree that software becoming worse (in some sense) but I don't blame AI.
My experiences as a user haven't really changed. Streamers fail to cast to TVs. Browsers are devices which reliably and efficiently propagate red console text and 500s from companies to you. I still see bluescreens on public touchscreens.
Uncle Bob nailed one reason why the industry isn't getting better: the growth of programmers is exponential, such that half of all programmers have less than n years experience (5?).
I offer two additional reasons:
Firstly, you hone your skills doing individual algorithms on a single thread, where it's OK to get it wrong and change it as many times as you like.
When you get a job, you don't even realise you're in a distributed system and the getter/setter model you practiced on is woefully inadequate for interacting with partner systems, cqrs, event-sourcing, auditability, idempotency or whatever else is required.
In short, 10 years of practicing writing for-loops and data-structures will not translate into building a system that customers can access via their phone and multiple tabs of a browser, at the same time as a database schema is being updated.
Secondly, while I don't think there is any realistic alternative to agile development, it seems that nontechnical POs have captured the software lifecycle. 'Focus on happy path only, for the MVP' is how you build buggy software. Good developers can learn that this will guarantee the need for a rewrite in the future, but POs never will.
It seems like the languages have been increasingly targeting the beginner segment (C++ -> Java -> JS -> Python), while the tasks have been getting more complicated (b2b distributed systems, 24/7 uptime, millions of users, resistance to hackers, machine learning, etc.)
> it seems that nontechnical POs have captured the software lifecycle
Yes, this dynamic has indeed been screwed up in many teams!
In my perspective, you can’t be a good product manager for a technical product without being somewhat technical yourself. You simply won’t understand your product well enough.
The only dynamic where that works is if the product manager acts more as a supporter and challenger to a team lead by an EM or technical lead, but then product manager is probably not the right title.
This is probably going to sound elitist, but it is how I've been seeing this whole LLM thing play out.
I don't necessarily believe in 10x devs, but there is a very clear output difference between various developers. During my career I've been both in a position where I was delivering much higher quality code at a similar pace to other coworkers and also mediocre code at a slower pace. Sometimes at the same time.
Now, AI itself is a force multiplier. Depending on the task, your setup, your skill level (both as a developer and with LLMs), the weather, the model and so on, it can range anywhere from actively hindering you, to solving the entire thing for you. Let's say that on average you get a 20% productivity boost.
If you take a 0.8x developer and give them a 1.2 force multiplier LLM, you get a 0.96x developer. Worse, the axis that has moved is speed, not quality.
All this to say that if you let a below average developer vibecode, you get a much higher volume of slop. You can make an agent create good code, but this developer didn't care before so why should they now?
*Note that this is from experience. I've started getting feature PRs from our devops team. It's been a very enlightening experience.
That’s why i‘m using claude to literally write everything for myself! I cannot fix software-problems in my car or my fridge or my heat-pump. But I can start by replacing my own infrastructure!
MacOS contstant trolling with tab switching. I regret each day of life I need to use this shitty OS because of corporate. In linux these things will never happen
Software has neither gotten worse nor become any better. All you're seeing is the cost of writing code approach near zero. As that happens the speed at which you see defects increases. Software is by its very definition "malleable". It is without (or very few) constraints. So it is very hard to get "prefect". That is why IMO Site Reliability Engineering will be one of the few fields that will remain, because until we build an _actual_ goodness to gracious Intelligent System that can think critically, including how systems can fail (which includes humans), software will be what is hat always been "soft".
> I opened Slack on macOS, the icon kept bouncing in the dock for a few seconds. I got impatient, switched to Ghostty, and started typing. Just then, the Slack window appeared, stole focus from Ghostty and the git pull command was sent to the group chat.
I still don't understand how a full focus stealability debounce is not the norm. If I am in the middle of writing, or interacting with one UI bit in any way, no other app should be able to steal focus, full stop. Add an annoying popup if you want, make angry computer noises, flash icons, I don't care. 99.9, if not 100% of the time, it's completely unrelated to what I'm doing and can absolutely wait. While some OSs have something along these lines, they also give the tools to bypass it (I'm looking at you, Cisco AnyConnect! Being connected after clicking on connect is not an emergency!)
Imagine you're working in a terminal and an app decides to steal STDIN. Why do we accept that for a GUI?
Yes! Few things piss me off more. It boggles the mind that this behavior is the norm, and it happens all the time! There are practically 0 situations in which I want something to automatically steal focus.
For some reason I have to deal with the opposite problem and have to click twice on a new window to get keyboard focus. Many times I'll switch windows (eg. from Messages to the browser), start typing and accidentally end up sending someone a message instead- even though the messages tab is completely out of view.
> I opened Slack on macOS, the icon kept bouncing in the dock for a few seconds. I got impatient, switched to Ghostty, and started typing. Just then, the Slack window appeared, stole focus from Ghostty and the git pull command was sent to the group chat.
One of my absolute favourite features on KDE Plasma with Wayland is the global setting to control what can steal focus. It works wonderfully, and I always miss it when I have to use my work mac or windows computers.
The single most soul-crushing experience I have on today's tech is the whoopee cushion.
I type this on a touch device that randomly Cheevers the words. (It literally changed "changes" just now. Henceforth, my word for this shall be Cheevers.) it Cheevers the words. It offers suggested words that you can tap to complete the word you're typing, and milliseconds before your finger makes contact with the screen, the suggested word flips to Cheevers and your finger taps that.
Now with the new improved iOS Liquid Glass, and possibly with janky web app development kits, buttons can move as you try to click on them. UI animation was supposed to help the user. But helping the user is hard.
Pop ups. The web needs more pop ups. More "Are you enjoying this app?" in your face.
But on the desktop, I recall complaining to a Windows (the OS) developer about stealing focus... in 1995. At the time, classic macOS tended not to do this, but it happened quite often multitasking on Windows, and I couldn't offer any constructive feedback, no obvious solution.
The sheer unpredictability, at 100 millisecond time scale... a few hours of this and I'm literally shaking. Because it affects muscle response, a startle reaction. Keeps firing away at the limbic system.
If I wanted slot-machine online gambling levels of simulation, I'd go to a casino. I hate this stuff. I hate where "engagement" metrics and the quest for more clicks has brought us.
On Windows, UWP was supposed to solve this (by blocking apps from taking focus directly, but having the OS/shell create an app window immediately on launch and give focus to it) but most apps aren't UWP apps.
My "favorite" aspect of the Slack web app, which might have been fixed, as I haven't seen it recently, is if you typed up -> r (I believe it's r) to open the emoji picker, typed a few characters (e.g. "eyes") and pressed enter. Well, if the emojis hadn't loaded by then (and apparently they're loaded fresh every time), they would pop in and often, the a different emoji would be picked than the one that was visibly highlighted when I pressed enter.
These are the small but extremely user-centric quality of life features that I like about KDE. Most applications I use aren't KDE based (Firefox, VSCode, emacs, movie) but KWin, Dolphin, Plasma panels have unmatched ease whih grows on you.
My Arch Linux KDE Desktop with the Haswell i7-4770S CPU (from 2013) actually feels snappier than MacOS on my MacBook Pro M4. Sure, the MacBook has far superior performance overall, but if you just want to move some windows or get a response from a right-click, it just feels slower.
179 comments
[ 3.1 ms ] story [ 79.0 ms ] threadThe smaller the org, the better the chances of having only reasonable levels of misalignment... but without taking on high risk, it's just better pat to work on a large, worse org. Frustrating.
Neither one really understands the other’s component.
Cooperation is almost impossible since such would derail one project manager’s unrealistic schedule and demean their authority.
If you're making a Linux distro, please don't do this. I will not watch your video. I was interested, and now I'm not.
Imo we know the tools (social, + technical) to make good software that works. But in the face of massive complexity, people just want the next feature, the next thing sales can use as a win
And yeah I agree with the other commenter here - I don't think this is AI. We had terrible software before, even declining quality.
I work in mediation and have recently seen a few cases where I’ve wondered if the participants were suffering from AI psychosis as related to their claims: where they’d constructed their legal cases hand-in-hand with an LLM and had become absolutely convinced despite a contestable set of facts.
People have always been subject to delusion, but in these cases, all documentation submitted was obviously created by chatbots.
It's glaringly obvious that poorly checked, AI-written, code is the culprit, I'm just wondering when companies will wake up.
Nobody gives a flying fuck about resiliency, good practices, security or any of that now that AI is there. All anyone cares about is pure output of features.
Any and all "productivity" gains attributed to AI efficiency are actually due to the complete disregard for any care or standard in the software industry in the past 2 years. Yeah you can output more shit when you don't give a fuck, big surprise, AI doesn't change anything here.
This comes back to the users. As long as they keep buying the shit software why spend money on QA.
I mean, you're the problem. Did you stop using slack? Are they taking losses because of it? If not, why change?
Optimistically, I think we might just enter a new phase at some point, where businesses make very conscious decisions about what to invest in to differentiate. Like, everyone has support bots now. Where do we invest, absolutely stunning human support? Or is support not that key and we should invest into being the most reliable? The most enjoyable to use?
One can hope.
So taking it at face value is kinda part of the problem here.
The only way this can get better is by completely suffocating it off attention. It doesn't exist. It's not real.
My experiences as a user haven't really changed. Streamers fail to cast to TVs. Browsers are devices which reliably and efficiently propagate red console text and 500s from companies to you. I still see bluescreens on public touchscreens.
Uncle Bob nailed one reason why the industry isn't getting better: the growth of programmers is exponential, such that half of all programmers have less than n years experience (5?).
I offer two additional reasons:
Firstly, you hone your skills doing individual algorithms on a single thread, where it's OK to get it wrong and change it as many times as you like.
When you get a job, you don't even realise you're in a distributed system and the getter/setter model you practiced on is woefully inadequate for interacting with partner systems, cqrs, event-sourcing, auditability, idempotency or whatever else is required.
In short, 10 years of practicing writing for-loops and data-structures will not translate into building a system that customers can access via their phone and multiple tabs of a browser, at the same time as a database schema is being updated.
Secondly, while I don't think there is any realistic alternative to agile development, it seems that nontechnical POs have captured the software lifecycle. 'Focus on happy path only, for the MVP' is how you build buggy software. Good developers can learn that this will guarantee the need for a rewrite in the future, but POs never will.
It seems like the languages have been increasingly targeting the beginner segment (C++ -> Java -> JS -> Python), while the tasks have been getting more complicated (b2b distributed systems, 24/7 uptime, millions of users, resistance to hackers, machine learning, etc.)
Yes, this dynamic has indeed been screwed up in many teams!
In my perspective, you can’t be a good product manager for a technical product without being somewhat technical yourself. You simply won’t understand your product well enough.
The only dynamic where that works is if the product manager acts more as a supporter and challenger to a team lead by an EM or technical lead, but then product manager is probably not the right title.
I don't necessarily believe in 10x devs, but there is a very clear output difference between various developers. During my career I've been both in a position where I was delivering much higher quality code at a similar pace to other coworkers and also mediocre code at a slower pace. Sometimes at the same time.
Now, AI itself is a force multiplier. Depending on the task, your setup, your skill level (both as a developer and with LLMs), the weather, the model and so on, it can range anywhere from actively hindering you, to solving the entire thing for you. Let's say that on average you get a 20% productivity boost.
If you take a 0.8x developer and give them a 1.2 force multiplier LLM, you get a 0.96x developer. Worse, the axis that has moved is speed, not quality.
All this to say that if you let a below average developer vibecode, you get a much higher volume of slop. You can make an agent create good code, but this developer didn't care before so why should they now?
*Note that this is from experience. I've started getting feature PRs from our devops team. It's been a very enlightening experience.
Microsoft didn't need AI to make windows 11 as bad as it is.
The incentives are large companies are fundamentally not in align with making high quality software people actually enjoy interacting with.
This is hilarious.
I still don't understand how a full focus stealability debounce is not the norm. If I am in the middle of writing, or interacting with one UI bit in any way, no other app should be able to steal focus, full stop. Add an annoying popup if you want, make angry computer noises, flash icons, I don't care. 99.9, if not 100% of the time, it's completely unrelated to what I'm doing and can absolutely wait. While some OSs have something along these lines, they also give the tools to bypass it (I'm looking at you, Cisco AnyConnect! Being connected after clicking on connect is not an emergency!)
Imagine you're working in a terminal and an app decides to steal STDIN. Why do we accept that for a GUI?
One of my absolute favourite features on KDE Plasma with Wayland is the global setting to control what can steal focus. It works wonderfully, and I always miss it when I have to use my work mac or windows computers.
See here for docs, under "Focus stealing prevention" - https://docs.kde.org/trunk_kf6/en/kwin/kcontrol/windowbehavi...
The single most soul-crushing experience I have on today's tech is the whoopee cushion.
I type this on a touch device that randomly Cheevers the words. (It literally changed "changes" just now. Henceforth, my word for this shall be Cheevers.) it Cheevers the words. It offers suggested words that you can tap to complete the word you're typing, and milliseconds before your finger makes contact with the screen, the suggested word flips to Cheevers and your finger taps that.
Now with the new improved iOS Liquid Glass, and possibly with janky web app development kits, buttons can move as you try to click on them. UI animation was supposed to help the user. But helping the user is hard.
Pop ups. The web needs more pop ups. More "Are you enjoying this app?" in your face.
But on the desktop, I recall complaining to a Windows (the OS) developer about stealing focus... in 1995. At the time, classic macOS tended not to do this, but it happened quite often multitasking on Windows, and I couldn't offer any constructive feedback, no obvious solution.
The sheer unpredictability, at 100 millisecond time scale... a few hours of this and I'm literally shaking. Because it affects muscle response, a startle reaction. Keeps firing away at the limbic system.
If I wanted slot-machine online gambling levels of simulation, I'd go to a casino. I hate this stuff. I hate where "engagement" metrics and the quest for more clicks has brought us.
Make it stop. I'm trying to work, here.
I think that points do the solution. We should not make stuff more complex than we can the level of complexity at which we can make it reliable.