Jason Gorman

Can anyone put their finger on when exactly the software industry stopped being about solving problems?

I think it's great how we all suddenly became capable of completely and precisely specifying software systems in natural languages after 7 decades of failing miserably at it.

Developers: "The problem is that the faster we go, the more we get things wrong and have to fix them or do them over again, which actually makes us slower overall"

Big Tech: "I see, yes. What if... hear us out... what if you could get things wrong much, much faster?"

After 34 years in software development, I'm beginning to wonder if being "left behind" might in fact be the secret to a long career.

"Across multiple coding agents and LLMs, we find that context files tend to reduce task success rates compared to providing no repository context, while also increasing inference cost by over 20%"

I've suspected this all along. Folks spending mucho-plenty time curating project-level .md files have been deluding themselves that it helps.

https://arxiv.org/abs/2602.11988

I'm treating myself to We Bought A Zoo

Good luck capturing decades of experience about software design in your .md files.

Picasso has a napkin with your name on it.

Hmm. When someone schedules a meeting without asking me when I'm available.

That's got that "beck and call" flavour to it.

Claude's at 2 nines reliability. Let's see if they can get that down to 1 nine by the end of the year.

Did you know the global software developer population's much *larger* than it was in 2022?

Did you know tech layoffs were in full swing when ChatGPT launched?

Did you know AI companies are hiring engineers in large numbers?

A quick fact-check:

https://codemanship.wordpress.com/2026/03/02/is-ai-killing-software-engineering-jobs/

Today, on LinkedIn Says The Funniest Things:

"Are people being paid to write comments and articles denouncing AI as a fad"

Yes, it's a conspiracy funded by Cory Doctorow, Gary Marcus and Ed Zitron. Or, as we know them collectively, "Big Blog".

Watching Canary Black. Computer expert looks at some code for about 3 seconds:

"It's a virus. Looks like it corrupts and encrypts all Internet traffic, locks down all devices connected to it."

internet.traffic.all.corrupt()
internet.traffic.all.encrypt()
connectedDevices.all.lockDown()

You must be using the Special Programming Language - the one that stops programmers making significant design choices.

I see posts on LinkedIn from people who believe dev teams should be pushed to adopt "agentic" coding so they can reap the "10x" productivity gains.

Putting aside the lack of evidence for even 1.5x gains on teams, if it really were possible, why would they need to be pushed towards it?

Fun factoid: when I led a team writing software for the British Standards Institute, they were *very* interested in our testing approach and asked me to write it up for them.

The software testing industry standards they manage were never mentioned.

Perhaps we could trick billionaires into funding public services by turning them into AI start-ups

There's a reason why businesses keep a human in the loop in many of their core processes. And it's the same reason I keep myself very much in the loop in "AI"-assisted programming.

https://codemanship.wordpress.com/2026/03/01/the-double-edged-sword-of-automation/

Bear with me on this. What if owning stuff isn't a skill, and people shouldn't get paid for it?

I learn so much about business from The Apprentice.

You design a product following the direction of the loudest person in the room.

And then, after the product can't be changed, you do market research.

How can I explain royalty to people in countries that don't have hereditary monarchies?

We tried selective breeding of humans, and it didn't work.

"Ah, but what is 'accuracy'?"

I refer the honourable gentleman to my previous answers about people who resort to philosophy in discussions of practical matters.

Just flipping through the Rolodex in my mind of "What if" scenarios relating to strategies built around LLMs - e.g., what if the hyper-scalers go bust? - and how the risks could be mitigated.

The only Plan B that would work in *every* scenario is "people".

It's becoming ever clearer that, far from leading to "super-intelligence", scaling LLMs is more likely to end at "super-mediocrity" - producing outputs that are closer to the average of their training data.

For software developers, this has implications. The average quality of publicly-available code examples is *not* good.

https://codemanship.wordpress.com/2026/02/25/super-mediocrity/

The decline in program comprehension and reasoning I've witnessed in heavy LLM users in recent months makes me think that they should spend significant time on a regular basis working unaided. It's like living in low gravity.

When you run out of tokens, carry on with the work, perhaps. If you *can't*, that should ring alarm bells.

My standard advice when devs express concern about "falling behind" when they read about the amazing things other people are doing with "AI" coding assistants is to look at their bio.

Are they calling themselves an "AI <something>"? Yes?

Then they're probably lying.

You realise that LLMs are Brawndo?

Software developers won't be replaced by AI. They'll be replaced by other developers using AI who don't care if the code works or even compiles working for bosses who don't care either.

Watch software developers. I mean *really* watch them.

When they say they spent an hour "writing code", they've actually spent most of that hour *reading* code and *thinking* about it.

Wanna boost your developer productivity?

* Write code that's easier to understand

* Get better at understanding code

More code ain't gonna help here.