CEOs who think AI replaces their employees are just bad CEOs

People think otherwise with AI partly because Anthropic kept telling us that they didn't have to write code or review code any more for most of their work. Their agent swarms just comb through their github, slack and wikis to figure out what to do next, and another swarm of agents just review, test, merge, deploy, A/B test, and revert the code. Boris alone merged nearly 300 PRs in the past week (or two?). So the top research labs seem have broken the productivity seal.

And then they talk about this recursively self-improving AI that is so powerful, so autonomous that they advocate that every company should be prepared to "pause" the effort. And their Fable/Mythos has this specific restriction as mentioned in their model card[1] that they are going to reject requests to tune and train models because, well you guess it, the models are too powerful to be used by mere mortals.

[1] We’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms. Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT).

If you can't make the visitor understand your entire complex product in 10 seconds, then you've lost them.

Your product has to be complex because that's where the software market is at. All of the low-hanging fruits have been taken by the time you identify them. Sure, someone will find a way to make money using new low-hanging fruits that arise due to technological changes but it's not going to be you. You probably don't have the business connections to make that work.

There is a definite lack of appreciation for the often repetitive grit, toil and maintenance work required to just have profit generating working software running reliably in production.

Don't think I've heard that one but certainly rings true to my experience.

Reminds me of "ninety percent of the game is half mental"

As a matter of fact I tend to think of it the first 80% is 80% of work, the 80% of the 20% left is same as first 80% of work. The last 4% polishing is another first 80% of the work. I think that is a good rule for general project management.

Then there's the technical debt!

Shipping is frankly the easy part. It's the operating overhead that often breaks you.

I liken it to free puppies.

It's as if the installation part is the hard bit, and after that it'll take care of itself for ... far enough into the future that it won't be <current manager>'s problem. It is solved.

... and that's just using the system, not fixing bugs and adding features.

I am not a children person. But I love this analogy.

To deliver something nice, we also must accept some suffering.

Good CEOs don't see their employees as an expense.

In that sense Altman and Dario et. al. were extremely successful in their cringeworthy campaign to establish themselves as priests of a new machine god religion. Even the people who don't want it believed them.

The good news is at least this year companies are starting to get a little more thoughtful about why they're paying for AI and what specific business function it serves.

Realities:

1. AI is a tool. You don't replace a job with a tool, just like you don't replace an apple with a sock. It was always an error of classification to think this way.

2. AI is a useful tool. For every CEO who thinks it justifies mass layoffs, there are dozens of people who don't want to admit that it does have a lot of utility. Anyone who isn't figuring out where this tool can make them more productive will have a hard time down the line imho.

3. We can infer from the priors that jobs will continue to be a thing, just like they still were after the inventions of printing presses, cars, power looms, etc. but there will be some pretty massive churn, some roles maybe disappearing entirely, new ones getting created, same goes for skill sets.

4. Businesses will use this churn to the best of their ability to either reduce costs or increase output. That second part is the key the AI haters don't seem to recognize. Thanks to competition, not every business wants to just fire everyone. Yeah, America's financialized and monopolized and less competitive than it used to be, but still, plenty of businesses will repurpose their budgets to produce more and expand.

So in terms of real specific and concrete things I've seen so far.

If you're in customer support that's a pretty rough spot, a well designed RAG system can turn 20 minutes of research into 20 seconds. Customer support budgets flow downstream from actual usage/customer base so if they can do more with less people will get laid off, they don't expand scope, they automate and cut the budget.

If you're a developer, your position's a lot more secure than that, coding agents are pretty incredible but they are simply not at a point where they can think through architectural decisions, and they occasionally go off the rails and trash everything. These things need to be operated and steered. Yes that's 100% the future of the profession and I'm sorry if someone doesn't like that, being a blacksmith who forges metal things by hand is also very niche these days, kind of sad if you loved blacksmithing but it happens. Devs also have to be aware that a PM/product owner can likely do some of their job now and I think any dev whose preference was to avoid thinking about customer or business requirements is going to find his utility is shrinking. A mass shrinking of the profession is unlikely but things will change and the only rational paths forward honestly are to retire or to figure out now how you fit in, it's a career opportunity if you get ahead of the curve even.

Lastly there is a whole other domain now which for lack of a good term I'll refer to as "prompt engineering," it requires some level of systems design thinking, but basically no programming expertise. The best candidate for this work is a person who possesses business process knowledge as well as systems oriented thinking. Maybe these jobs will end up finding a home in the IT department or something. As an economy we're barely recognizing them yet but I see that in engineering we can increasingly implement places for a prompt input, and then hand some workflow/business domain decisions off to someone who understands them better and they just tweak their prompts over time to get a great system. Pretty sure new jobs will be created and formalized around this over time.

I think it’s exceedingly rare that a CEO is actually competent at their job. In most cases it’s the labor class propping the company up, and in some cases the workers are doing so against the wishes of the CEO. Not that executives want to ruin the company, they’re just incompetent and therefore make terrible decisions constantly.

on the contrary, it seems to be one of the few jobs that seems to require absolutely no qualifications to have.

What you need to do to be CEO is.... convince someone to lend you money in the hope that you'll get it back to them.

I've worked under some absolutely awful people who wouldn't pass an interview anywhere, but somehow they're CEOs, because they can smarm there way into more money consistently.

Most of them got into a prestigious school on legacy, paid for by wealthy parents. Many were above average IQ, but by no means geniuses. They had access to computers earlier than others, due to said affluence. They seem unable or unwilling to comprehend they're overwhelmingly on average, "nepo babies" to steal a term from the world of entertainment.

With that said, I've been programming for 25 years and I've only been a CEO for 3, so take what I said with a pinch of salt.

I do think people overestimate titles like this a lot, though, and it really comes down to what the company actually does and what is demanding for that company at that position/role. The CTO of a some-bullshit-as-a-service company may as well be straight out of college, because they're likely doing something trivial that literally anyone (including LLMs) could put together. The CTO of a well-used and reliable streaming service that handles a meaningful part of the world's Internet traffic is obviously solving a more interesting and demanding problem, and their decisions are going to be more important.

There was a time when I used to recommend "Out of the Crisis", a book from Demming, to business leaders.

Problem is, "Out of the crisis" still assumes as a premise that companies compete on the quality of their products, that making money comes from actually making and selling stuff. That leaders are not the anti-intelectual morons that believe that absolutely any thing can be explained with a 15 minutes deck, that math and statistics are passtimes for weird geeks that don't add "business value" and because of that, is a book that could have steered us toward a better world in the 80s, but now it is completely useless, because its recipes can't handle the level of degradation things goto into.

It literally just requires filing an LLC or Corporation. There are several SaaS companies that will do it for you.

The Open AI guy said it's now better than doctors (he said it, not me lol). He's replaced his doctor, right?

CEOs get signals from multiple channels that cantbe accessed by AI or ordinary workers

so that alone becomes a moat

same with sales jobs theres a plane that AI cannot observe (yet)

The code is human + AI, the management is only AI

(discussed 6 months ago https://news.ycombinator.com/item?id=46072002)

Would you rather take instructions from a ruthless robot or ruthless flesh sack?

Naive and stupid, but it was downvoted and flagged away into oblivion with zero chance for a conversation.

This assumes that a mass consumer economy is necessary, when it isn't. Mass consumption is relatively new, for most of history economies functioned with just a small consuming elite and large underclass that consumed very little. We are already approaching that again in the states given that the top 10% of earners are already responsible for nearly half of all consumer spending.

There's a floor even in a mostly automated economy where some services are resistant to automation simply because the human element is the product. Luxury hospitality, personal care, etc. That billionaire is going to want a human masseuse, not a robot.

A highly automated economy could stabilize like this with a small elite population consuming luxury goods & services, served by a low-wage economic underclass human workforce.

Its certainly not a pleasant society, but its also not unsustainable given enough oppression or pacification (bread and circuses anyone?)

I wrote about it here too:

https://blog.nilesh.io/post/llms-and-jobs

Just because people have a vague idea of what someone's job is and a deep understanding of their own job, it feels like AI can replace everyone but you.

L2 issues are already involved in some way often revealing some kind of system failure, requiring context and exploration to understand, and judgement (and perhaps even system overrides) to fix.

I could see “automated L2 is the new L1” improvements, but without a big capability jump and/or a resource bonfire, I don’t think even frontier models would effectively replace good L2 staff.

They might magnify good L2 staff so fewer are needed (and maybe even help L1 staff become L2).

You know who can't do that? People who call L1 support.

"AI psychosis" tries to group together the phenomenon where people fall in love with an LLM-generated character, the phenomenon where people spend too much time talking philosophy with LLMs and stop expressing coherent thoughts, and the phenomenon where tech evangelists say "AI can do all of these 5 million things" when actually it can't do all of them. But what if these are all different phenomena with different causes and solutions?

Which indicates: the management believes there are productivity gains from AI use, but adoption lags due to inertia and reluctance to change existing workflows.

I call bull on that. Employees often have just as much stress, often imposed on them from those higher up, where if they don't do well they could get fired and it could mean they can't put food on the table.

Several times I've worked in environments where the stress was incredibly high, and once it even put me in the hospital after a long stretch of working 70 hour weeks to satisfy the insane deadlines my CEO boss wanted (this was at a startup).

Whereas if an executive screws up they just golden parachute to another cushy gig elsewhere, oftentimes, because even if they screw up, they're still valued as having experience running companies by other companies. And even if they cannot, they still likely got paid enough that they won't be struggling financially if they go through a period without a job.

Depends. Is the robot looking out for our best interests? Or is the robot looking out for the shareholders?

I already ask AI to manage a lot of things, including my own activities. It all depends on whether the model can be trusted to act in our best interests.

Ideally contractors that benefit you personally (eg: your buddy who now owes you one), but definitely contractors that let you outsource the responsibility.

Even better if you get some management consultant to suggest the idea and/or do the subcontracting.

Definitely buys you a few quarters of bonus and some time to land your next gig.

You mean, this is an entirely made-up figure.

For who? Will it ever be read? I’m bias on good documentation and I think AI is better at publishing documents for machines, for now.