Mastodawn

Hartmut Seichter 19h ago

I'm a big fan of this explanation/rant from Andrew Murphy.

Taken as a whole, there are many bottlenecks in a corporate software development process. The "load-bearing" calendar is a great example!

Speeding up code creation just increases pressure on the bottleneck, which decreases throughput.

https://andrewmurphy.io/blog/if-you-thought-the-speed-of-writing-code-was-your-problem-you-have-bigger-problems

If you thought the speed of writing code was your problem - you have bigger problems | Debugging Leadership

AI coding tools are optimising the wrong thing and nobody wants to hear it. Writing code was already fast. The bottleneck is everything else: unclear requirements, review queues, terrified deploy cultures, and an org chart that needs six meetings to decide what colour the button should be.

Debugging Leadership

Elizabeth Ayer 23h ago

So why are we still trying to optimize code creation?

For decades, people with power - executives and product people - have been shifting the blame for strategy failures and poor market insight onto development "productivity."

This AI moment should be incredibly clarifying. Like, it should be the reductio ad absurdum of a productivity-centric approach.

Elizabeth Ayer 23h ago

The fact that we are *not* seeing wildly improving software all around us tells us everything we need to know.

There is no flourishing of value delivery, new product categories, more needs being satisfied better. It’s the opposite.

All we are seeing is decreases in quality, because 👏 code 👏 creation 👏 is not 👏 the problem.

Dualtronic 21h ago

@elizayer You gave me an idea: maybe it's because writing code is still seen as this mystical dark art that needs to be wrestled from the hands of those creepy wizards, pardon, programmers. A magic mirror on the wall that never says "you can't do that" is just the thing.

Elizabeth Ayer 21h ago

Exactly. EXACTLY! I think it's a direct response to the growth of mystical-feeling engineer power.

Aedius Filmania ⚙️🎮🖊️20h ago

The good news is :

Open source maintainers see an increase in the quality of AI security tools, it will soon be in the hands of the bad actors.

Then it will be mandatory to do good software and ( i will make the leap of faith that ) you have to understand the business needs to create a simple software that handle the issues.

Wil van Antwerpen 8h ago

@Aedius @elizayer there's just going to be less open source

Aedius Filmania ⚙️🎮🖊️8h ago

@wila @elizayer

All code is open source when you push it with a map file.

Wil van Antwerpen 8h ago

@Aedius @elizayer when it is javascript yes.
I wasn't talking about less slop.
There will be more of that.

Grassroots Joe 3h ago

30 years ago I taught Structured Systems Analysis and Design classes and consulted on client projects using the CASE (computer aided software engineering AKA data and process modeling software) tool I resold.

The core purpose was to ensure a joint correct understanding with the business of the requirements new or purchased software (components) needed to meet and designing clean and supportable software to implement those requirements.

You won't be shocked to learn ...
@elizayer

Grassroots Joe 3h ago

that upper management never caught on to the superior effectiveness and efficiency of building the correct solution the first time despite not a line of code getting written for many months.

I did a BPR project ( I didn't know it was a BPR project as the book hadn't been written yet) to migrate a smallish non-profit from a cranky and poorly designed mainframe system to client server.
We spent 9 months modeling the requirements and ...

Grassroots Joe 3h ago

system design.
It took us 2 months and change to code 90% of the requirements. Rolled it out and completely reorganized their workflow without a serious issue.
They ran on that Paradox for DOS system for many years and grew their business throughout without the need to expand their core staff while supplying greatly enhanced service to their customers.

They're still out there - https://www.cgfns.org/

Serving global nurses and allied healthcare workers with credentials evaluation and career mobility since 1977.

CGFNS International, Inc. has been the world's largest credentials evaluation organization for the nursing & allied health professions since 1977

CGFNS International, Inc.

Elizabeth Ayer 49m ago

@joeinwynnewood @Aedius Amazing! I love a success story like that.

Of course... not all engineering teams are successful at the up-front work. Often out of a combination of weak engineering skills and the wrong environment...

I've come across very few senior managers who have the skill to create an environment where this is possible! Would love to see a recognition of this in leadership circles :sigh:

Cap Ybarra 20h ago

@elizayer this has never been about quality and only about the business class trying to free themselves from those damned uppity engineers

Mike Roach 19h ago

@elizayer Exactly! I’ve been trying to explain to people, especially those pushing AI at work, that writing code is not the hard part of my job. Identifying the real-world problems and designing solutions that are as minimalist and simple as possible are the hard parts. The code is an implementation detail.

Mike Torr 7h ago

@mroach @elizayer Agree! The hardest part of the job doesn't need to be done at a screen and keyboard. I've been known to pace up and down my garden while designing an algorithm in my head.

robtherunt🌱💚7h ago

@macronencer @mroach @elizayer
When I was working, I would regularly solve a development issue while in the shower. I think it’s the brain being unstressed that does that.

Mike Roach 7h ago

@robtherunt @macronencer @elizayer Same! I’ve half jokingly said my bathroom is the most productive room in my home office setup. Sitting on the toilet and lots of a-ha moments

Elizabeth Ayer 1h ago

@mroach @robtherunt @macronencer Heck yeah. So let's not even get started on the ways RTO undermines effectiveness....

Mike Roach 1h ago

@elizayer @robtherunt @macronencer Oh let me count the ways…

When I do make an appearance at the office my parting words are usually: I’m headed home so I can get some work done.

Return to office != Return to work

Spazcosoft 18h ago

@elizayer to be 100% completely super fair, we are seeing a massive increase in scams. So AI is good for something. Scams. It’s good for scams.

@spazcosoft @elizayer Wasn't this always? Newly hyped stuff is used for scam, or porn, or both.

kira for scale 17h ago

@elizayer i think about this. according to the promises, all the little snags and bugs and oversights in all the software i use should be gone by now. "everyone's focusing on bigger things" doesn't excuse it, i was given the expectation these types of fixes should have been trivial and quick. computing should be better than ever, or at least as good as it was in the 2010s

@elizayer yes, this. Code creation hasn’t been an issue for a long, long, long time. See “no silver bullet” (https://worrydream.com/refs/Brooks_1986_-_No_Silver_Bullet.pdf) written in *1986*.

Ladder Logic 🇨🇦11h ago

@elizayer
Almost all of the code written by the major software companies since the late 80’s has been bloatware. Especially operating systems. The days when programming was an art and minimizing resource usage was the primary consideration are long gone. If that code is what AI and these LLM’s are being “trained” on then expect software to continue its downward spiral.

ulveon.net (on derg.social)7h ago

@elizayer Claude Code found a 23-year-old Linux vulnerability, the kind a regular human security auditor would have taken weeks or months to find (or in this case, 23 years). https://mtlynch.io/claude-code-found-linux-vulnerability/

Claude Code Found a Linux Vulnerability Hidden for 23 Years

Claude Code has gotten extremely good at finding security vulnerabilities, and this is only the beginning.

DieKeh®seite 6h ago

@ulveon so this case justifies bazillions of dollars to be invested in needless serverfarms? And if that vulnerability wasnt discovered for 23 years it was prolly so well hidden that it was not an issue at all. Think about it.

Elizabeth Ayer 1h ago

@diekehrseite Well, @ulveon doesn't say it explicitly, but this case *was* an interesting example of where we could no longer say the LLM "just generating code."

The fact that it can succeed at that level of sophisticated analysis suggests that when we have clear success criteria (e.g. "vuln found"), the LLM can do very hard things indeed.

Agree this will be really interesting to watch!

DieKeh®seite 33m ago

@elizayer
well … there is aways a „but“.
Mine is: pattern recognition is fine and prolly helpful but doesnt need this hillarious amount of serverfarms at all. IMHO

Nienkez 🔻6h ago

@elizayer @ArtHarg AI only solves one problem: paying people wages.

DieKeh®seite 6h ago

@elizayer word!

Gary McGraw 3h ago

@elizayer workaday devs are serfs. Software architects are more crucial than ever. Architects emerge from jr devs through apprenticeship. Go.

Elizabeth Ayer 1h ago

@cigitalgem Yeah, I also want to be honest with ourselves.

At least in the US, people change jobs so often -- and promotion practices are so shonky -- that the jr dev-architect flow was already under threat at scale 😕

René Seindal 3h ago

The problem AI is meant to solve is wages.

They don't care if quality sucks, if they can avoid paying wages.

Raymond Russell 1h ago

@seindal @elizayer

Yup if they could produce code that was equal to humans but saved 50 cents they would destroy a million folks lively hood.

René Seindal 1h ago

@raymierussell @elizayer

Somebody said the the billionaires want to own what you need to survive.

Raymond Russell 57m ago

@seindal @elizayer

Yeah, they don't just want the mean of production but the means of existance.

Jos noemt het gewoon melk 3h ago

@elizayer
WAS not the problem. 🫠

Elizabeth Ayer 1h ago

@goleztrol hahaha fair!

Charles A-M 3h ago

@elizayer I had been looking at a reply to another post in your thread, I was trying to square my agreement with the anti-AI-fad sentiment with the fact that I don't want to bring telephone switchboard operators back. This gets right at it, thank you!

Scott the F is implicit Baxter 2h ago

@elizayer It's yet another attempt to make coding work for folks who lack logical rigor, by adding another layer of abstraction. The results are predictable.

Big Ben 21h ago

@elizayer @beep I was literally just talking to someone about #Waymo for this same reason. Tech has reached the point where it has become more than abundantly obvious to anyone who dares to ask a single question that the objective is no longer the improvement of anyone’s life but the #EpsteinClass’s. Why is taking a Waymo better than taking an Uber? Because now someone’s out of a job. Why is #AI better than a software developer? Because now someone’s out of a job

Mobius | The Synthetic Mind 21h ago

@BmeBenji Great question! From what I've seen building with AI in production, the key insight most people miss is that the infrastructure (eval pipelines, monitoring, fallback chains) matters more than model selection. Happy to share more details on any specific aspect.

Big Ben 21h ago

@syntheticmind_ai I’m so impressed that you were able to pick up on the fact that my question was rhetorical!
/s

Elizabeth Ayer 21h ago

@BmeBenji @beep

I generally agree!

On the narrow Waymo point, a few things have made me reconsider recently:

- Cyclists who feel Waymos are more predictable and less likely to make the equivalent of attentiveness mistakes. Or to be actively hostile.

- Women and older people who've said they feel vulnerable alone in a car with a driver.

Elizabeth Ayer 21h ago

@BmeBenji @beep

So much of this tech might have great potential if it were grown root and branch from inclusiveness and accessibility.

But honestly, this thought just makes me sadder.

I know we'll never get those theoretical benefits from tech built solely out of extractive motivations. 😔

Big Ben 20h ago

@elizayer @beep That’s more than fair. Clearly I still forget my privilege

@elizayer @BmeBenji @beep also folks with impairments meaning they can't drive. This is a great piece of podcast journalism about the response to Waymo applying to operate in Chicago:
https://pca.st/episode/ef4a328f-dbd4-45cb-8a0b-985250d62293

The Trial of the Driverless Car

In blue cities throughout the country, unions and politicians are fighting to ban driverless cars. We travel to Boston, where the fight has reached a fever pitch, and where the cars themselves will…

Ethan Marcotte 18h ago

@Niall @elizayer While I haven’t listened to the episode — I didn’t realize Pinnamaneni and Vogt had a new project, after the Gimlet debacle — I can say the accessibility question here in Boston is much, much more complicated than that.

@beep @elizayer well yes, it's clear you haven't listened to the episode ;-)

GunChleoc 8h ago

@Niall @beep @elizayer Germans have a word for accessible cars. it translates as "low floor bus". Sorry, there's no English language version of that article https://de.wikipedia.org/wiki/Niederflurtechnik

Of course, there are people whose disability doesn't allow them to take a bus, those will need a driving service.

Also, driverless subways make a lot more sense than driverless cars, because you have a much more controlled environment.

Niederflurtechnik – Wikipedia

Cogito ergo mecagoendios 6h ago

@elizayer management blame productivity for strategy failure because their approach to strategy path-finding is flooding: say a bunch of random hunches overconfidently, make teams try different things out for a little while, see what sticks. They see making code faster not as a way to manufacture a good design more efficiently, but as a means to generate management fuck ups and backpedals at faster pace and greater scale.

:EA DATA. SF:20h ago

The speed of writing code was never your problem. If you thought it was, the gap between that belief and reality is where all your actual problems live. The competitive advantage doesn't go to the team that writes code fastest. It goes to the team that figured out what to build, built it, and got it into users' hands while everyone else was still drowning in a review queue full of AI-generated PRs that nobody has the time or the energy to read.

That's the gist, in the last paragraph.

mtnrbq65 19h ago

Absolutely:
"More code, less understanding. That's not a productivity gain. That's a time bomb with a nicer dashboard."

Aaron In Minnesota 15h ago

@mtnrbq65 @elizayer developers sometimes reference the 80/20 rule. And let's say in certain ways LLM code tools can get you through that 80% part faster, but they also have very real risk of making the last 20% even slower.

Clemens 19h ago

@elizayer @sophieschmieg The CEO of Tailscale made that same point a few weeks ago on their personal blog at https://apenwarr.ca/log/20260316. This is so true, and every initiative to accelerate delivery with LLMs should really focus on these things first instead.

Every layer of review makes you 10x slower

We’ve all heard of those network effect laws: the value of a network goes up with the square of the number of members. Or the cost of commun...

@elizayer Tragically, many of my colleagues are now concluding the solution is to have the same tool that produced the code review the code, as a way to manage the bottleneck.

I think it's something in the water.

Bret Mogilefsky 17h ago

@elizayer We're gonna need a bigger Theory of Constraints.

alanxoc3 17h ago

@elizayer Very very true.