"We’re heading off a cliff if we continue to put AI first and humans last." Well said Jim. So much of this also applies to the increasing reliance on generative AI in the creative industries.
How will AI give you everything for free? AI companies expect you to buy their tokens - nothing is going to be free. Also, how is AI going to the most critical work that runs civilization when it's proven to be so innaccurate and incompetant. The hype about what it can do is just lies and fraud from nefarious corporations run by nihihilstic techlords.
People need to be more honest with each and other and start living in the real world. In the quest for unlimited money, lots of devs/firms are completely forgetting about security first principles, data protection and data privacy. It's madness. When I read yesterday Brian Armstrong's statement that in his firm (Coinbase) people were shipping code through vibecoding I was shocked, literally. The illusion is so powerful and dangerous that some are completely detached from reality.
Maybe after the crash and dust all settles, software design and engineering skills will come back in style? Besides the decline in software quality that AI-first development entails, avoiding slop token costs or copywriter concerns may prove to be the better and more profitable overall option for skilled teams. Or maybe I'm just trying to remain hopeful in dystopia lol 🙃
I've been wondering if software development was always going to be an ephemeral occupation. Automotive assembly line workers, lamplighters, typesetters, elevator operators, and many others, they were always going to be replaced--they just didn't know it at the time.
I've written software for more than 35 years, now. I believe it's time for human written software to die. It will die, you know, it may take some years, but its days are numbered.
Maybe. Which begs the question: what do we do now if we're middle-aged but not ready to retire? And what about an entire generation or two behind us who were highly encouraged to pursue the same career only to have the rug pulled on them?
When I asked Claude for the list, he gave some that lasted only a decade or less. 60 years for software sits in the middle. Draftsmen and typesetters were longer lived. Are our skills transferrable? Are those target occupations also slated for demolition? I queried Claude some more, he was not optimistic.
Very strange response. Is this an ad hominem attack? You don't seem to believe in anything. Nihilist? This doesn't contribute to the conversation at all.
What aspect doesn’t contribute? You seem to equate nihilism with realism.
Your two questions above are based on unwarranted assumptions. First, it addresses the “plight” that anyone at any age is owed anything. Second, those younger generations can figure it out, unless you have absolutely no faith in their critical thinking abilities or their ability to grow and pivot.
Your original issue operates on assumptions that were fallacies to begin with.
So again, how does this not contribute?
What is curious is why anyone would be foolish enough to trust any system, much less any career path as a guarantee of anything. Your comments here clearly make assumptions that have no logical place in the discussion.
In your responses to other writings I see you’ve been critical of AI and and the AI industry. But here you decided to posit “So what? Pull yourself up by your bootstraps and quit complaining, stupid humans.” Which just seems very out of place. There has never been this much of a concerted effort by the very rich and powerful to end all human knowledge work. Yet you seem okay just dismissing the threat and blaming individual workers. I think you must be very young and haven’t experienced a layoff yet? What will you do when your own career ladder completely disappears? You assume pivoting to something else while everything else is also under threat is somehow easy. It definitely won’t be!
I think your premise is not accurate. First of all, automotive workers are still here, and the attempt to replace them fully has failed (Tesla tried). Lamplighters are still here too, they are now electricians. Elevator operators are still here, they are now elevator mechanics. Typesetters are still here, they are now digital designers. Heck, horse carriage drivers are still here, they are now taxi and Uber drivers. For each of these examples the job hasn't really gone away, as the human need for the service still exists, but it has shifted in responsibility, the specific tasks, and the skills required. Interestingly enough, every time the world has seen a new technology to automate tasks, more jobs were created than before, to the point where today the total number of jobs is larger than ever in known human history.
As for software engineering, the job has shifted many times before. The technology has completely changed since the job as first created, the levels of abstractions we deal with on a daily basis has grown to an almost unbounded degree, and the productivity in economic terms of any single developer has grown to previously unimaginable levels, judging by the salary levels we’ve seen and are still seeing. I don’t think AI will replace what is at the core of the software engineering profession, which is to solve problems in some (semi-)automated, reliable way. Or in less lofty terms, get things done using technology.
The specific task responsibilities might shift, the tools used will change, and specific skills required will shift alongside. However the core ability to identify and solve problems will remain in demand.
"Yes, by late 2025 into early 2026, reports indicated that China successfully resurfaced or built a 158 km stretch of highway (often cited as part of the Beijing-Hong Kong-Macao highway) using AI-powered robots, autonomous pavers, and drones coordinated over a 5G network, requiring zero human workers physically present on the road surface."
Guess we don't need all those orange-vested human statues standing around a road job.
Hmm, did I claim that the people had died away or the occupations? Software development, as an occupation, will fade away. The humans? Maybe they can make pottery, or deliver fast food, music? Nope, that's getting nixed, too. I know, we'll all become plumbers.
The goal is for software to write itself. Then they will focus on humanoid robots to replace most manual labor. The goal is to make humans redundant. Some of these tech oligarchs believe in Extinctionism: the complete usurpation of humans by robots. That's their religion and we should understand that as their motivation.
I’m critical of systems with arbitrary barriers to entry, I’m opposed to the idea of being at the mercy of or reliant on any organization. I’ve never trusted or relied on any org, to have any individuals best interest in mind, and foolish of anyone to put any stock in org structure, or tenure as any basis for anything. I’ve also sought to use any tools at my disposal to retain how I do things, augment how I do things, and reduce the need to focus on craft if I can achieve the same thing my own way. I also believe most orgs are too large, and that we should not need enterprises or institutions to exist. Further, the more individuals can function independently without actually needing a team to execute work, while also being able to create systems to execute the work independently of other silos is a net good.
I’m not critical of the tech. I’m critical of the orgs, and that they won’t actually allow users to utilize the full scope of the tech.
I see things now, more so that individuals like you and the other individual, could in theory have depth of domain that could be leveraged to own more of your own knowledge and institutional expertise.
And here’s my point and where I grow frustrated. There are folks like you who could design and architect BETTER systems for leveraging this tech for the good of others and those who are vulnerable, or those who feel let down by the system. Yet, the posts and articles you have criticizing emerging tech with a broad strokes, fails to account for those who can leverage it well, because we already had systems for semi automating the executions of work, and now can extend that more through how we might leverage the tech.
I’m critical of what I see in the markets, not the tech and where it’s been leveraged well.
I guess I don't see how AI makes us better. It's too probabilistic and hallucinatory to be reliable, and the only real money being made so far is by unscrupulous chatbot companies preying on minors, deepfake criminals, and AI-washing wrapper startups. I don't see any substantive advantage to outsourcing our cognition or creativity. I'm not ignorant: I've explored the tools and understand how they work, and I don't think they offer solutions to any real problems except in a few niche scenarios. For me, there is a gigantic chasm between the hype and the reality, and I think we're obsessing over this at the neglect of everything else that matters.
So then why not build and architect solutions that can give individuals the option? I don’t really see it as outsourcing cognition or creativity, because that assumes orgs where these are leveraged were ever actually allowing or enabling individuals to do this, and creativity assumes no instrument.
I’d argue, it enables offloading low value tasks, to allow more creativity and augmented cognition. I’d also argue that your point doesn’t account for those who are neurodivergent, or don’t fit within broken org structures and refuse to waste time with things like admin overhead and politics, etc.
I get to focus more on creative problem solving, because I’m not wasting my energy on executive functions that I biologically do not have. So, the assumption that working memory and processing speed are a viable measure of anything is really a moot point now, and that’s a fantastic thing for anyone. Unless you are someone who values things like rote memorization and modern day schooling which doesn’t actually teach creativity or critical thinking, or abstract reasoning.
Can we have both? Can we design systems that are optimal for neurodivergent individuals but not push an explictly dehumanizing tech on everyone? My fight is mostly against forced mandates, ignorance of the risks to our world and our psychology, and belief in the myth that AI is some inevitable, unstoppable force. I just want to promote coming back down to reality. You can use AI as a tool but I want to see an end to proselytizing and using AI as a weapon or a way to oppress people and accumulate wealth for the 1%. I feel like a lot of people in this debate are like single issue voters who refuse to look at the problem as a whole.
Then it seems we agree on the root issue, because quite literally every point you mention here, is where I’ve been for the past 8 years, and longer for my full career.
Where we diverge, is that I personally began designing around all the issues, and now have the opportunity to design and architect better than what we see and have now.
And, while I have verbally avoided discussing this, but that’s the entire gap around what I’ve seen, and wanted to have better than what we have to pay for in the market… plus… none of them ever have options for a majority of the world. And given my hands on, technical, experience, and so forth, anything out there should be for the user, period, and not extractive or exploitative, and should be designed with all that by default at the foundation… and most would never consider building half of what’s possible, because it goes against the value extraction ops model… which is also the root cause of all the other issues in orgs.
And if nothing else, every person should own their intelligence. not be at the mercy of enterprise software.
I always like to go back to the classics: “The Psychology of Computer Programming” by Gerald Weinberg. “The Mythical Man-Month” by Frederick Brooks Jr, and his 16-page PDF “No Silver Bullet” (The “AI” section, where he quotes David Parnas, is brilliant and relevant 40 years later… Brooks wasn’t awarded a Turing for nothing…) - “The Joel Test” is aging, but it’s generally excellent. I also like ramblings like Ron Jeffries’ columns “It’s Not AI” and “Story Points Revisited”. Even work-adjacent books like “Peopleware” by Tom DeMarco and Tim Lister. There’s probably a lot more.
This isn't the first time some marketing-driven managerial craze has swept through the c-suite of software firms. It won't be the last. People who manage software but don't know how it, or the production of it, works, are incredibly vulnerable to snake oil salesmen. The pain points are real: it's expensive, hard to predict timelines, hard to scale teams, prone to unforeseen failure.
Every time some quack comes along with plausible-sounding claims that he can cure what ails software management, the answer is "I'll take the entire batch."
I was here for "Object Oriented Programming fixes everything!"
I was here for "SCRUM fixes everything!"
For kanban and pomodoro and pair programming, for TDD and memory safe languages and continuous integration, for functional programming and reactive programming and containerization.
Now it's agentic programming. Now the snake oil salesmen have found the cajones to market based on the suspicion quietly harbored by the nontechnical managers (and the rest of the company) the entire time: the problem is not your processes and policies, which are thorough; it's not the tools or resources, on which you've spent more than enough; nor is it your ideas, which in the words of Gemini, are "brilliant insights!," or your business models, which exceed industry standards.
The problem, you see, is what you always thought it was: your developers are slow, lazy, insubordinate, close-minded, insufferably tactless, nitpicky, obsessive compulsive, anal retentive, and above all, *human*. But with this one weird trick, for the low low price of $15 per million tokens (wow! So many tokens!) we can fix everything.
Thanks for writing this up. Reflects well with experiences I see.
But I'm wondering whether this is all due to AI ... software engineers working on their own, preferrably out of their home offices to do "their thing", is a trend I see since Covid. Having leadership that doesn't want to hear critics from their team, product owners who don't measure what success means is a trend that has been here before the rise of AI Coding assistants. (At least from what I can see.)
But even if AI would be just an amplifier of things that were already going wrong - I agree that we "have to sound some kind of alarm that work culture is being eroded".
Yes I’ve been considering how the tendency for developers to want to work alone and avoid coworkers and collaboration as a possible catalyst or a reason why some devs are embracing AI more than others. I have an article in the works where I try to explore this idea.
I think that physical proximity makes a difference. AI Chatbots in your home office can deliver content, but not such proximity. It's an important factor, but undervaluedby many in my eyes.
This nails the corporate side. But there is a mirror story you are not seeing.
While enterprises are drowning in failing AI mandates, individual builders are using the same tools to ship things that would have taken whole teams a few years ago. Not for OKRs. Not for a CTO. Just because the thing deserved to exist. I've been solo-building an open source ecosystem simulation engine, no runway, no regrets.
And I'm not an edge case. People are building weird and beautiful, uncommercial software everywhere right now. Some of them are broke. They are building anyway!
All the anguish about flattened career ladders and broken rubrics only hurts if you are still chasing the Porsche (or whatever is the thing now) and the Pacific Heights condo. There's another path, the bohemian one. Musicians, painters and writers have lived frugally and made work that mattered for centuries. Software just got so captured by compensation culture that we forgot it could be done any other way. Now the tools exist to do it modestly and some of us are choosing that trade. Give up the status symbols annd get your mind back. Live light and build what obsesses you and trust that life tends to take care of people who are fully immersed in making something real.
Your Kafkaesque fever dream is real but it's a corporate phenomenon. Outside the org chart, what's happening looks more like a pre-cambrian explosion, a chaotic blooming of creative forms that simply couldn't have existed before, built by people who need no one's permission to create.
The craft you're mourning inside the enterprise isn't dead. It just left the building...
I respect the solo creators, but do passion projects pay the mortgage and provide your family with healthcare? I've turned my back on corporate America but I'm not ready to retire so I'm now living quite precariously and facing a lot of uncertainty. Unfortinately I don't think most people can or should choose this path. There's no safety net.
Jim, fair point and I wont romanticize it. The precarious nature is real and honestly the consulting fallback that used to cushion this kind of leap is more saturated than we have ever seen it. Everyone who got laid off is chasing the same contracts!
But thats kind of the point. The old deal is breaking from the inside too. The safety net you are describing ( the corporate one ) is the same one thats shredding people right now. So we are all precarious. The only question is what you end up holding when the music stops.
Bohemians never had a clean answer to the mortgage question. They just organized life differently. Kept the overhead radically low. Pieced it together teaching, freelance, commissions they didn't love to fund the work they did. I think Picasso was broke for years. Jazz musicians gigged six nights and made the real music on the next night. It was never comfortable. It was just honest about what it cost.
Software engineers (or just how the economy and culture is setup here and the induced stress. Another topic!) never had to learn any of this because the compensation was so absurd for so long. Now that its normalizing, might be worth knowing the playbook exists, not because it's easy but because the alternative isn't working either. Rooting for you out there.
This connects to a structural shift that hit this week. Anthropic and OpenAI just announced near-identical JVs with Blackstone, Goldman, TPG, and Bain to ship forward-deployed engineers into PE portfolio companies. The unit of measurement is no longer 'employee impact' or 'team impact,' it's 'how much cost structure did the model plus the embedded engineer remove.' That math eats the engineering manager role faster than it eats the engineer.
I think the key to understand here is that the builders of AI are expressly trying to replace human intelligence, while at the same time admitting that they do not understand human intelligence. I think we're in the frothy stage like the first dotcom bubble when the public conversation was driven by people in marketing, not in engineering. In the long run, I think AI will be more like a prosthetic that enhances human capacities than a drop-in replacement for human beings.
I do agree with you that the narrative, "We're going to replace all human labor in three years," is in sharp conflict with the reality, "We need people who have money to purchase our tokens." Those two ideas just can't exist together. And for some stupid reason, nobody is talking about it.
Jim, thoughtful piece as always. I appreciate your insistence on human first thinking.
However, the typical engineering manager (like I was) under layers of corporate hierarchy is not in a position to push back in any meaningful way. YMMV, especially if you are further up the chain.
The general disposition of upper management, in my experience, is that everyone is replaceable except themselves, and they are quick to label anyone who dissents as not a team player.
Furthermore, the way tech industry (and our economy in general) is structured these days gives incumbents almost insurmountable leverage, eliminating meaningful competition. So what do they care if the product quality goes down, as long as their stocks go up?
Upper management are rarely held accountable, even if they make all the decisions that lead to total collapse. This has been the case for decades before AI.
Seems to me it's a game where the only way to win is not to play. Let the whole stinking heap collapse and then maybe, just maybe, there'll be room for a new beginning.
Way to kill my one shred of optimism lol. I know. I know. I sometimes revert back to idealism. I can't help it. It's too depressing to give up on humanity even though that's in vogue right now.
I'm not giving up on humanity, I'm giving up on the current economic model where wealth inevitably concentrates at an accelerating rate. That has nothing to do with AI and everything to do with our culture and economic model.
Mathematics is beautiful, algorithms are a subtle subset, but engineering is not a craft. It is a process.
Having worked in software engineering for nearly 50 years, one thing is clear. Software is entirely disposable, and software engineers love to think otherwise.
The horror of AI is to see both how trivial most “code” is (a machine could write it), and the fact that we could advance in a few years - it is 2026, right? - from nothing to better than human.
Better? Having worked in software engineering for 50 years (around the world!) most people working in “software engineering” are barely adequate. The Peter Principle arrives very quickly when you systematically test work products.
Here’s the retreat:
It can’t write code.
It writes bad code.
It can duplicate the best algorithms for tasks since it’s trained on them and humans aren’t but it’s not original.
Handwritten code has nuance a machine can’t duplicate.
It’s not code, it’s data structure. AI is poor at organizing data.
It can organize objects but not architect data.
It can’t organize functions.
It can organize collection of functions, but it doesn’t really architect.
It can’t architect a systems.
It’s bad at architecting systems.
It can’t organize a project to deliver complex software.
It can’t organize a project, architect, or delivery code in a way which is signifiant.
AI will never write commercial software. Nobody will buy AI written software.
AI can’t replace the team, the QA, governance, the A B C D
Lather rinse repeat.
AI constructed software is a fact resulting from billions of lines of code doing almost exactly the same thing over and over fed to a model. You just don’t need people to write the same invoice processing or order entry code, sorting files or sending a payload to an API anymore. Render, transform, orchestrate, callback,
Defend your ground! All the way up to the top! But if we’re here in just a few years, oh no.
Daft Punk gave the most cogent analysis.
Buy it, use it, break it, fix it, trash it, change it, mail, upgrade it Charge it, point it, zoom it, press it, snap it, work it, quick, erase it Write it, cut it, paste it, save it, load it, check it, quick, rewrite it Plug it, play it, burn it, rip it, drag it, drop it, zip, unzip it. Lock it, fill it, call it, find it, view it, code it, jam, unlock it Surf it, scroll it, pause it, click it, cross it, crack it, switch, update it Name it, read it, tune it, print it, scan it, send it, fax, rename it Touch it, bring it, pay it, watch it, turn it, leave it, stop, format it
"We’re heading off a cliff if we continue to put AI first and humans last." Well said Jim. So much of this also applies to the increasing reliance on generative AI in the creative industries.
You seem bitter. What is your personal experience of the creative industry?
Well yeah, that's capitalism. But it's the only system we've got.
How will AI give you everything for free? AI companies expect you to buy their tokens - nothing is going to be free. Also, how is AI going to the most critical work that runs civilization when it's proven to be so innaccurate and incompetant. The hype about what it can do is just lies and fraud from nefarious corporations run by nihihilstic techlords.
People need to be more honest with each and other and start living in the real world. In the quest for unlimited money, lots of devs/firms are completely forgetting about security first principles, data protection and data privacy. It's madness. When I read yesterday Brian Armstrong's statement that in his firm (Coinbase) people were shipping code through vibecoding I was shocked, literally. The illusion is so powerful and dangerous that some are completely detached from reality.
Yeah I read that too. Sheer insanity. I'm genuinely worried about how all this will end up. This is a mental health epidemic.
Very true.
Maybe after the crash and dust all settles, software design and engineering skills will come back in style? Besides the decline in software quality that AI-first development entails, avoiding slop token costs or copywriter concerns may prove to be the better and more profitable overall option for skilled teams. Or maybe I'm just trying to remain hopeful in dystopia lol 🙃
I do think IF people come to their senses, those of us who still know how to use our own brains will become a prized commodity.
We already are.
AI is turning all of life into a Kafkaesque fever dream for everyone.
I've been wondering if software development was always going to be an ephemeral occupation. Automotive assembly line workers, lamplighters, typesetters, elevator operators, and many others, they were always going to be replaced--they just didn't know it at the time.
I've written software for more than 35 years, now. I believe it's time for human written software to die. It will die, you know, it may take some years, but its days are numbered.
Maybe. Which begs the question: what do we do now if we're middle-aged but not ready to retire? And what about an entire generation or two behind us who were highly encouraged to pursue the same career only to have the rug pulled on them?
When I asked Claude for the list, he gave some that lasted only a decade or less. 60 years for software sits in the middle. Draftsmen and typesetters were longer lived. Are our skills transferrable? Are those target occupations also slated for demolition? I queried Claude some more, he was not optimistic.
LOL kind of shitty of Claude not to even offer a slither of hope for us.
Maybe try having an identity and sense of self outside one’s career or trade or place within an org?
Is there an assumption based on arbitrary entitlement that any person is guaranteed anything simply because they put forth effort?
Unclear if this has ever been the case? Not sure the markets reward effort, or guarantee “dream fulfillment” unless history has some evidence here?
Seems like a personal problem, that an individual should resolve.
Very strange response. Is this an ad hominem attack? You don't seem to believe in anything. Nihilist? This doesn't contribute to the conversation at all.
What aspect doesn’t contribute? You seem to equate nihilism with realism.
Your two questions above are based on unwarranted assumptions. First, it addresses the “plight” that anyone at any age is owed anything. Second, those younger generations can figure it out, unless you have absolutely no faith in their critical thinking abilities or their ability to grow and pivot.
Your original issue operates on assumptions that were fallacies to begin with.
So again, how does this not contribute?
What is curious is why anyone would be foolish enough to trust any system, much less any career path as a guarantee of anything. Your comments here clearly make assumptions that have no logical place in the discussion.
In your responses to other writings I see you’ve been critical of AI and and the AI industry. But here you decided to posit “So what? Pull yourself up by your bootstraps and quit complaining, stupid humans.” Which just seems very out of place. There has never been this much of a concerted effort by the very rich and powerful to end all human knowledge work. Yet you seem okay just dismissing the threat and blaming individual workers. I think you must be very young and haven’t experienced a layoff yet? What will you do when your own career ladder completely disappears? You assume pivoting to something else while everything else is also under threat is somehow easy. It definitely won’t be!
I think your premise is not accurate. First of all, automotive workers are still here, and the attempt to replace them fully has failed (Tesla tried). Lamplighters are still here too, they are now electricians. Elevator operators are still here, they are now elevator mechanics. Typesetters are still here, they are now digital designers. Heck, horse carriage drivers are still here, they are now taxi and Uber drivers. For each of these examples the job hasn't really gone away, as the human need for the service still exists, but it has shifted in responsibility, the specific tasks, and the skills required. Interestingly enough, every time the world has seen a new technology to automate tasks, more jobs were created than before, to the point where today the total number of jobs is larger than ever in known human history.
As for software engineering, the job has shifted many times before. The technology has completely changed since the job as first created, the levels of abstractions we deal with on a daily basis has grown to an almost unbounded degree, and the productivity in economic terms of any single developer has grown to previously unimaginable levels, judging by the salary levels we’ve seen and are still seeing. I don’t think AI will replace what is at the core of the software engineering profession, which is to solve problems in some (semi-)automated, reliable way. Or in less lofty terms, get things done using technology.
The specific task responsibilities might shift, the tools used will change, and specific skills required will shift alongside. However the core ability to identify and solve problems will remain in demand.
"Yes, by late 2025 into early 2026, reports indicated that China successfully resurfaced or built a 158 km stretch of highway (often cited as part of the Beijing-Hong Kong-Macao highway) using AI-powered robots, autonomous pavers, and drones coordinated over a 5G network, requiring zero human workers physically present on the road surface."
Guess we don't need all those orange-vested human statues standing around a road job.
Hmm, did I claim that the people had died away or the occupations? Software development, as an occupation, will fade away. The humans? Maybe they can make pottery, or deliver fast food, music? Nope, that's getting nixed, too. I know, we'll all become plumbers.
Who will develop the software if not software developers, whatever we call that profession in the future?
The goal is for software to write itself. Then they will focus on humanoid robots to replace most manual labor. The goal is to make humans redundant. Some of these tech oligarchs believe in Extinctionism: the complete usurpation of humans by robots. That's their religion and we should understand that as their motivation.
Except of course the software won't write itself. On robotics I agree.
The software already DOES write itself.
All the LLMs are using iterative self-improvement. And that's just today. Today will be the worst AIs compared to all the tomorrows.
I’m critical of systems with arbitrary barriers to entry, I’m opposed to the idea of being at the mercy of or reliant on any organization. I’ve never trusted or relied on any org, to have any individuals best interest in mind, and foolish of anyone to put any stock in org structure, or tenure as any basis for anything. I’ve also sought to use any tools at my disposal to retain how I do things, augment how I do things, and reduce the need to focus on craft if I can achieve the same thing my own way. I also believe most orgs are too large, and that we should not need enterprises or institutions to exist. Further, the more individuals can function independently without actually needing a team to execute work, while also being able to create systems to execute the work independently of other silos is a net good.
I’m not critical of the tech. I’m critical of the orgs, and that they won’t actually allow users to utilize the full scope of the tech.
I see things now, more so that individuals like you and the other individual, could in theory have depth of domain that could be leveraged to own more of your own knowledge and institutional expertise.
And here’s my point and where I grow frustrated. There are folks like you who could design and architect BETTER systems for leveraging this tech for the good of others and those who are vulnerable, or those who feel let down by the system. Yet, the posts and articles you have criticizing emerging tech with a broad strokes, fails to account for those who can leverage it well, because we already had systems for semi automating the executions of work, and now can extend that more through how we might leverage the tech.
I’m critical of what I see in the markets, not the tech and where it’s been leveraged well.
I guess I don't see how AI makes us better. It's too probabilistic and hallucinatory to be reliable, and the only real money being made so far is by unscrupulous chatbot companies preying on minors, deepfake criminals, and AI-washing wrapper startups. I don't see any substantive advantage to outsourcing our cognition or creativity. I'm not ignorant: I've explored the tools and understand how they work, and I don't think they offer solutions to any real problems except in a few niche scenarios. For me, there is a gigantic chasm between the hype and the reality, and I think we're obsessing over this at the neglect of everything else that matters.
So then why not build and architect solutions that can give individuals the option? I don’t really see it as outsourcing cognition or creativity, because that assumes orgs where these are leveraged were ever actually allowing or enabling individuals to do this, and creativity assumes no instrument.
I’d argue, it enables offloading low value tasks, to allow more creativity and augmented cognition. I’d also argue that your point doesn’t account for those who are neurodivergent, or don’t fit within broken org structures and refuse to waste time with things like admin overhead and politics, etc.
I get to focus more on creative problem solving, because I’m not wasting my energy on executive functions that I biologically do not have. So, the assumption that working memory and processing speed are a viable measure of anything is really a moot point now, and that’s a fantastic thing for anyone. Unless you are someone who values things like rote memorization and modern day schooling which doesn’t actually teach creativity or critical thinking, or abstract reasoning.
Can we have both? Can we design systems that are optimal for neurodivergent individuals but not push an explictly dehumanizing tech on everyone? My fight is mostly against forced mandates, ignorance of the risks to our world and our psychology, and belief in the myth that AI is some inevitable, unstoppable force. I just want to promote coming back down to reality. You can use AI as a tool but I want to see an end to proselytizing and using AI as a weapon or a way to oppress people and accumulate wealth for the 1%. I feel like a lot of people in this debate are like single issue voters who refuse to look at the problem as a whole.
Then it seems we agree on the root issue, because quite literally every point you mention here, is where I’ve been for the past 8 years, and longer for my full career.
Where we diverge, is that I personally began designing around all the issues, and now have the opportunity to design and architect better than what we see and have now.
And, while I have verbally avoided discussing this, but that’s the entire gap around what I’ve seen, and wanted to have better than what we have to pay for in the market… plus… none of them ever have options for a majority of the world. And given my hands on, technical, experience, and so forth, anything out there should be for the user, period, and not extractive or exploitative, and should be designed with all that by default at the foundation… and most would never consider building half of what’s possible, because it goes against the value extraction ops model… which is also the root cause of all the other issues in orgs.
And if nothing else, every person should own their intelligence. not be at the mercy of enterprise software.
For everyone watching, this is how adults engage and debate and argue. Which I will always respect.
I always like to go back to the classics: “The Psychology of Computer Programming” by Gerald Weinberg. “The Mythical Man-Month” by Frederick Brooks Jr, and his 16-page PDF “No Silver Bullet” (The “AI” section, where he quotes David Parnas, is brilliant and relevant 40 years later… Brooks wasn’t awarded a Turing for nothing…) - “The Joel Test” is aging, but it’s generally excellent. I also like ramblings like Ron Jeffries’ columns “It’s Not AI” and “Story Points Revisited”. Even work-adjacent books like “Peopleware” by Tom DeMarco and Tim Lister. There’s probably a lot more.
This isn't the first time some marketing-driven managerial craze has swept through the c-suite of software firms. It won't be the last. People who manage software but don't know how it, or the production of it, works, are incredibly vulnerable to snake oil salesmen. The pain points are real: it's expensive, hard to predict timelines, hard to scale teams, prone to unforeseen failure.
Every time some quack comes along with plausible-sounding claims that he can cure what ails software management, the answer is "I'll take the entire batch."
I was here for "Object Oriented Programming fixes everything!"
I was here for "SCRUM fixes everything!"
For kanban and pomodoro and pair programming, for TDD and memory safe languages and continuous integration, for functional programming and reactive programming and containerization.
Now it's agentic programming. Now the snake oil salesmen have found the cajones to market based on the suspicion quietly harbored by the nontechnical managers (and the rest of the company) the entire time: the problem is not your processes and policies, which are thorough; it's not the tools or resources, on which you've spent more than enough; nor is it your ideas, which in the words of Gemini, are "brilliant insights!," or your business models, which exceed industry standards.
The problem, you see, is what you always thought it was: your developers are slow, lazy, insubordinate, close-minded, insufferably tactless, nitpicky, obsessive compulsive, anal retentive, and above all, *human*. But with this one weird trick, for the low low price of $15 per million tokens (wow! So many tokens!) we can fix everything.
Thanks for writing this up. Reflects well with experiences I see.
But I'm wondering whether this is all due to AI ... software engineers working on their own, preferrably out of their home offices to do "their thing", is a trend I see since Covid. Having leadership that doesn't want to hear critics from their team, product owners who don't measure what success means is a trend that has been here before the rise of AI Coding assistants. (At least from what I can see.)
But even if AI would be just an amplifier of things that were already going wrong - I agree that we "have to sound some kind of alarm that work culture is being eroded".
Yes I’ve been considering how the tendency for developers to want to work alone and avoid coworkers and collaboration as a possible catalyst or a reason why some devs are embracing AI more than others. I have an article in the works where I try to explore this idea.
Nice, looking forward to read your thoughts.
I think that physical proximity makes a difference. AI Chatbots in your home office can deliver content, but not such proximity. It's an important factor, but undervaluedby many in my eyes.
This nails the corporate side. But there is a mirror story you are not seeing.
While enterprises are drowning in failing AI mandates, individual builders are using the same tools to ship things that would have taken whole teams a few years ago. Not for OKRs. Not for a CTO. Just because the thing deserved to exist. I've been solo-building an open source ecosystem simulation engine, no runway, no regrets.
And I'm not an edge case. People are building weird and beautiful, uncommercial software everywhere right now. Some of them are broke. They are building anyway!
All the anguish about flattened career ladders and broken rubrics only hurts if you are still chasing the Porsche (or whatever is the thing now) and the Pacific Heights condo. There's another path, the bohemian one. Musicians, painters and writers have lived frugally and made work that mattered for centuries. Software just got so captured by compensation culture that we forgot it could be done any other way. Now the tools exist to do it modestly and some of us are choosing that trade. Give up the status symbols annd get your mind back. Live light and build what obsesses you and trust that life tends to take care of people who are fully immersed in making something real.
Your Kafkaesque fever dream is real but it's a corporate phenomenon. Outside the org chart, what's happening looks more like a pre-cambrian explosion, a chaotic blooming of creative forms that simply couldn't have existed before, built by people who need no one's permission to create.
The craft you're mourning inside the enterprise isn't dead. It just left the building...
I respect the solo creators, but do passion projects pay the mortgage and provide your family with healthcare? I've turned my back on corporate America but I'm not ready to retire so I'm now living quite precariously and facing a lot of uncertainty. Unfortinately I don't think most people can or should choose this path. There's no safety net.
Jim, fair point and I wont romanticize it. The precarious nature is real and honestly the consulting fallback that used to cushion this kind of leap is more saturated than we have ever seen it. Everyone who got laid off is chasing the same contracts!
But thats kind of the point. The old deal is breaking from the inside too. The safety net you are describing ( the corporate one ) is the same one thats shredding people right now. So we are all precarious. The only question is what you end up holding when the music stops.
Bohemians never had a clean answer to the mortgage question. They just organized life differently. Kept the overhead radically low. Pieced it together teaching, freelance, commissions they didn't love to fund the work they did. I think Picasso was broke for years. Jazz musicians gigged six nights and made the real music on the next night. It was never comfortable. It was just honest about what it cost.
Software engineers (or just how the economy and culture is setup here and the induced stress. Another topic!) never had to learn any of this because the compensation was so absurd for so long. Now that its normalizing, might be worth knowing the playbook exists, not because it's easy but because the alternative isn't working either. Rooting for you out there.
Companies are not tech shops and never should be.
This connects to a structural shift that hit this week. Anthropic and OpenAI just announced near-identical JVs with Blackstone, Goldman, TPG, and Bain to ship forward-deployed engineers into PE portfolio companies. The unit of measurement is no longer 'employee impact' or 'team impact,' it's 'how much cost structure did the model plus the embedded engineer remove.' That math eats the engineering manager role faster than it eats the engineer.
https://systemshaywire.substack.com/p/the-day-ai-burned-my-brain-out
I think the key to understand here is that the builders of AI are expressly trying to replace human intelligence, while at the same time admitting that they do not understand human intelligence. I think we're in the frothy stage like the first dotcom bubble when the public conversation was driven by people in marketing, not in engineering. In the long run, I think AI will be more like a prosthetic that enhances human capacities than a drop-in replacement for human beings.
I guess we'll see.
I do agree with you that the narrative, "We're going to replace all human labor in three years," is in sharp conflict with the reality, "We need people who have money to purchase our tokens." Those two ideas just can't exist together. And for some stupid reason, nobody is talking about it.
Jim, thoughtful piece as always. I appreciate your insistence on human first thinking.
However, the typical engineering manager (like I was) under layers of corporate hierarchy is not in a position to push back in any meaningful way. YMMV, especially if you are further up the chain.
The general disposition of upper management, in my experience, is that everyone is replaceable except themselves, and they are quick to label anyone who dissents as not a team player.
Furthermore, the way tech industry (and our economy in general) is structured these days gives incumbents almost insurmountable leverage, eliminating meaningful competition. So what do they care if the product quality goes down, as long as their stocks go up?
Upper management are rarely held accountable, even if they make all the decisions that lead to total collapse. This has been the case for decades before AI.
Seems to me it's a game where the only way to win is not to play. Let the whole stinking heap collapse and then maybe, just maybe, there'll be room for a new beginning.
Way to kill my one shred of optimism lol. I know. I know. I sometimes revert back to idealism. I can't help it. It's too depressing to give up on humanity even though that's in vogue right now.
"He's dead, Jim" 🦴
I'm not giving up on humanity, I'm giving up on the current economic model where wealth inevitably concentrates at an accelerating rate. That has nothing to do with AI and everything to do with our culture and economic model.
the binary still matters for the leaders who haven’t picked a side.
for those who’ve moved past it, the integration question turns out to be harder than the framing.
Mathematics is beautiful, algorithms are a subtle subset, but engineering is not a craft. It is a process.
Having worked in software engineering for nearly 50 years, one thing is clear. Software is entirely disposable, and software engineers love to think otherwise.
The horror of AI is to see both how trivial most “code” is (a machine could write it), and the fact that we could advance in a few years - it is 2026, right? - from nothing to better than human.
Better? Having worked in software engineering for 50 years (around the world!) most people working in “software engineering” are barely adequate. The Peter Principle arrives very quickly when you systematically test work products.
Here’s the retreat:
It can’t write code.
It writes bad code.
It can duplicate the best algorithms for tasks since it’s trained on them and humans aren’t but it’s not original.
Handwritten code has nuance a machine can’t duplicate.
It’s not code, it’s data structure. AI is poor at organizing data.
It can organize objects but not architect data.
It can’t organize functions.
It can organize collection of functions, but it doesn’t really architect.
It can’t architect a systems.
It’s bad at architecting systems.
It can’t organize a project to deliver complex software.
It can’t organize a project, architect, or delivery code in a way which is signifiant.
AI will never write commercial software. Nobody will buy AI written software.
AI can’t replace the team, the QA, governance, the A B C D
Lather rinse repeat.
AI constructed software is a fact resulting from billions of lines of code doing almost exactly the same thing over and over fed to a model. You just don’t need people to write the same invoice processing or order entry code, sorting files or sending a payload to an API anymore. Render, transform, orchestrate, callback,
Defend your ground! All the way up to the top! But if we’re here in just a few years, oh no.
Daft Punk gave the most cogent analysis.
Buy it, use it, break it, fix it, trash it, change it, mail, upgrade it Charge it, point it, zoom it, press it, snap it, work it, quick, erase it Write it, cut it, paste it, save it, load it, check it, quick, rewrite it Plug it, play it, burn it, rip it, drag it, drop it, zip, unzip it. Lock it, fill it, call it, find it, view it, code it, jam, unlock it Surf it, scroll it, pause it, click it, cross it, crack it, switch, update it Name it, read it, tune it, print it, scan it, send it, fax, rename it Touch it, bring it, pay it, watch it, turn it, leave it, stop, format it
Over and over. Software engineering.
It’s being compressed.
That’s all.