> The real reason not to become an EM in 2026 is because AI makes our jobs 10x harder.

This is true, but our job was getting kind of boring anyway. Time to lead, not manage. We should be having just as much fun as the ICs, and the best I know are having the time of their lives.

While I too am only seeing a boost on the order of 20% so far, I think there are more creative applications of LLM beyond writing code, that can unlock multiples of net productivity in delivering product end to end. People are discovering these today and blogging about them, but the noise about dark factories and agents supervising agents supervising agents, etc, is drowning out their voices.

Every one of us is a pioneer if we choose to be. We have only scratched the surface as an industry.

Yes exactly, of all the uses cases for LLMs "writing code" is easily my least favorite. Theres so many other cool things for "stochastic contextual orchestrators"

Allegations of security theater should start with discussing the threat model. This is just somebody complaining about a crappy key card system.

To be fair, he was pointing out that the invisible "credentials in cookies" issue was much harder to get fixed:

The turnstiles were visible. They were expensive. They disrupted everyone's day and made headlines in company-wide emails. Management could point to them and say that we're taking security seriously. Meanwhile, thousands of employees had their Jira credentials stored in cookies. A vulnerability that could expose our entire project management system. But that fix required documentation, vendor approval, a month of convincing people it mattered. A whole lot of begging.

Again, not security theater. Signs of general dysfunction yes. Embarrassing. Fun to tease about for sure.

Aside: the more times I re-read the article the more annoyed I am with the self-righteous tone. It feels like the author is mimicking the style of legendary Usenet posts, but the story just isn’t that interesting and the writing not that witty, it falls flat.

If it isn't outright fake it's at least embellished. It even has the "and then everyone clapped" line!

The writing is clearly AI-generated or at least AI-assisted, so I think it's safe to assume it's also a work of fiction.

I’ll take your word for that. I don’t know how to tell. But I did notice that the writing was conspicuously terrible throughout. Entire sentences make no sense, such as “I'd slip in suspiciously while they contemplated the email that clearly said not to let anyone in with your own card.”

Turnstiles aren't theater and Redis doesn't make password storage secure so the entire thing seems a little el-el-emish..

But what about that sentence does that not make sense? They are describing tailgating..

It doesn’t make sense as a whole. But, for example, what was he suspicious of?

"I'd slip in suspiciously" means the "slipping in" was suspicious.

It can mean either. "Suspicious behavior" doesn't mean that the behavior thinks that you've done something wrong.

"She's suspicious" can mean either that I suspect her intentions or that she suspects someone else's intentions.

The last two paragraphs are mainly what stood out. I've spent hours trying to get LLMs to stop writing like that. It's hard because you can't just say things like "don't write lists of three items" because sometimes you want a list of three items. The rest of the text could be written by a person as it's kind of disjointed, but that could also be the result of trying to prompt out the AI-isms.

Just wanted to point out that you may be arguing for the article.. IE US style management is heavily inspired by Drucker and resistant to Deming.

Tests are not free, over proliferation of AI-touched tests is itself a problem, similar to the problem duplicative and verbose AI-generated code.

And tests are inherently imperfect, they may not test the perfect layer, so they break when they shouldn't, and they certainly don't capture every premise.

I'm on board with the tactics you suggest, but they are only incrementally helpful. What we really need is AI that removes duplicative code and unnecessary tests.

Destroying the inventory has a cost though.

Setting up the production line is the expensive part, firing things off it once it’s running is what’s cheap.

Storing stock is very expensive too.

This incentivizes what’s happening if you’re making cheap clothes.

I love this article but don't understand the conclusion. Heroku is dead as a doornail, of course.

Salesforce's core product was on bare metal up to a couple years ago. What they should have done is adopt Heroku as their internal Platform-as-a-Service. That would have solved three problems: 1) provided a ready and proven foundation for cloud adoption by Salesforce business units, 2) stimulated Heroku's product roadmap by giving it a very large and loyal design partner, and 3) eliminated the opportunity cost in terms of headcount, developer productivity, and poor imitation that came with the alternative "Falcon" aka "Hyperforce" project that became Salesforce's albatross and black hole for developer energy and goodwill going on 7+ years now.

> 2) stimulated Heroku's product roadmap by giving it a very large and loyal design partner

This is very much a double-edged sword. I've seen products get killed because they had one outsized customer with outsized influence over the product design and made it too specific to that customer rather than building something for everyone the customer would have to adapt to.

If they had, heroku would be very different today, since they aren't even doing enterprise contracts anymore (from what I saw of some other comments here). Maybe that would have been a good thing, maybe not.

> What they should have done is adopt Heroku as their internal Platform-as-a-Service

From what I saw, Heroku was unsuitable for a serious large company. Deploy-on-push is a nice UX for a small company, but once you need something more structured, it wasn't enough.

Can you elaborate on your claim it’s not enough? PasS (platform as a service) has been transformative for modern infrastructure. Containers for better, or worse, have made deployments substantially “easier,” giving every org the ability to provide Heroku like services. But, Heroku provided this sort of ease, and more, long before containers ate the world. And regardless of containers themselves, I fail to see how something “more structured” would even exist. Deploy on push is still subject to being merged and everywhere I’ve worked has had no shortage of checks and approvals to merge.

That lacks imagination. There's no corporate workflow that can't ultimately translate to moving git tags.

Perhaps, but why should one do it when it's a bad model ? Jury because it's possible ?

How did you objectively decide it's a bad model?

Because it's more enterprise to open a Service Now ticket and have Joe from IT upload the new content using FTP.

I subjectively did based on my experience.

Except you want to track what went to production and what didn't and for how long

Releases of a Heroku app is tracked by the platform, and both shown the the UI, in the CLI and available in the API.

Git is remarkably good at tracking history. CI systems are also great at showing history.

I previously deployed stuff to Salesforce when I ran a very large Asia-Pacific Salesforce org.

The previous way (prior to SFDX which was clearly influenced by Heroku) was terrible. 12 hour long deploys that end when one unit test times out-style terrible. No code history terrible. There is no way that Heroku was worse for integration.

Whether they could have replaced APEX with Heroku is a different issue.

It's based on this paper BTW: https://www.anthropic.com/research/small-samples-poison

Thank you! That's a fascinating paper.

It’s more than just money, it’s how you set up your life to be resilient to contingencies. For example finding a compatible life partner. For example finding happiness without lifestyle inflation and breaking free from the hedonic treadmill. Or perhaps having a good lifestyle business for some people. Or having extended family support nearby. I call these things unfuckwithability. Money is a big part of it, but may not be the biggest missing piece for many people.

My PS4 Slim was not capable of this at the device level. An individual app could choose to expose the choice of audio format, but many do not :(