Nostr notes by Jason Gorman

"what used to make someone a 'senior developer' — ...

2026-07-15T19:30:34Z

"what used to make someone a 'senior developer' — knowing syntax cold, writing boilerplate fast, remembering every API by heart — is being commoditized in real time"

RE: https://mastodon.cloud/@jasongorman/116844067988976085 Next ...

2026-07-01T11:53:12Z

RE: https://mastodon.cloud/@jasongorman/116844067988976085

Next week's Code Craft course is the final one for the knowable future. I don't think there's another training course quite like it.

Over the past 17 years, in one form or another, this training has changed more than a few careers.

2 spots left. Bring the boss a shiny apple before you ask them about it 🙂

Details in profile.

quoting
note1fkx…urs5

The Design Principles part of the next week's 3-day Code Craft workshop mostly focuses on something the majority of dev teams/orgs and a surprising number of architects are *really* bad at - modular design.

The exercise will task pairs with THINKING ABOUT WHAT SHOULD GO WHERE in a test-driven/BDD process, driving high-level modular designs from the outside in.

Yes, thinking about design is allowed. Just enough to keep your ship away from the rocks. And to get it off the rocks if it's too late.

The Design Principles part of the next week's 3-day Code ...

2026-07-01T09:43:06Z

The Design Principles part of the next week's 3-day Code Craft workshop mostly focuses on something the majority of dev teams/orgs and a surprising number of architects are *really* bad at - modular design.

The exercise will task pairs with THINKING ABOUT WHAT SHOULD GO WHERE in a test-driven/BDD process, driving high-level modular designs from the outside in.

Yes, thinking about design is allowed. Just enough to keep your ship away from the rocks. And to get it off the rocks if it's too late.

Research finds an interesting correlation between confidence in ...

2026-06-30T04:30:38Z

Research finds an interesting correlation between confidence in AI output and belief in the paranormal.

Another correlation I'm seeing with belief in the paranormal is the use of rhetorical tricks to flip the burden of evidence on to the person asking for it.

The biggest datasets from outfits who are in a position to ...

2026-06-29T06:53:22Z

The biggest datasets from outfits who are in a position to collect hard data from real dev teams, like Faros and CircleCI, show beyond any reasonable doubt that the *average* team that uses AI for code generation is outperformed in software delivery by the average team that doesn't.

True story.

I'm proud to know trans people who aren't hiding, and ...

2026-06-25T21:06:24Z

I'm proud to know trans people who aren't hiding, and I'll stand with them

"wanted to reach out because we're working with other ...

2026-06-23T12:32:31Z

"wanted to reach out because we're working with other trainings that are offering training workshops.‍

We're help trainings with targeted campaigns that are getting their sales teams meetings"

In some organisations, it just doesn't matter what dev teams ...

2026-06-23T04:23:45Z

In reply to nevent1q…r0c4
_________________________

In some organisations, it just doesn't matter what dev teams do, and they can absorb a *lot* of *big* failures - like the IT disasters we read about in the public sector, except theirs rarely make the news.

(2/2)

There was a time when I didn't feel the need to debate the ...

2026-06-23T04:23:33Z

There was a time when I didn't feel the need to debate the most wrongheaded, risky and harmful takes on software development because reality would eventually make the case for me.

I underestimated the capacity of some organisations to live with catastrophically bad outcomes if their moat's deep enough. (And, contrary to what you may read, most big business moats are things like property portfolios or rigged markets - not actual products or services.)

(1/2)

The growing trend for AI-accelerated Big Design Up-Front is ...

2026-06-20T05:31:49Z

The growing trend for AI-accelerated Big Design Up-Front is founded on a category error about the nature of the design process.

And faster BDUF isn't going to help. Quite the reverse, in fact.

https://codemanship.wordpress.com/2026/06/20/the-solution-to-bduf-isnt-faster-bduf/

Ship With Confidence Without Getting Stuck In PR Traffic Jams or ...

2026-06-10T08:30:45Z

Ship With Confidence Without Getting Stuck In PR Traffic Jams or Causing Pile-Ups In Production Using This One Weird Trick!

Last Codemanship Code Craft workshop - possibly ever? - July 7-9.

Only 3 places left.

https://codemanship.co.uk/codecraft.html

A take I see often online is that it doesn't matter if code ...

2026-06-07T07:42:59Z

A take I see often online is that it doesn't matter if code is "clean" - readable, simple, decoupled, easy & fast to test - as long as it works.

They're missing the point. Code that's hard to understand, complex, tightly-coupled and hard/slow to test is very likely to break.

There's no tradeoff.

But even if you could make quality gates good enough that you ...

2026-06-07T06:21:24Z

In reply to nevent1q…ljk7
_________________________

But even if you *could* make quality gates good enough that you won't need to review the code - and good luck with that - you then run straight into another trap: comprehension debt.

LLMs will never be reliable enough that you won't need to understand the code. That's physics.

"It's simple. All we need to do is agree on exactly what ...

2026-06-07T05:12:30Z

"It's simple. All we need to do is agree on exactly what we mean by 'bad code' and then codify all of that - including all the tradeoffs and value judgements informed by decades of experience but that we've never articulated and that are highly context-dependent - in a computable form that runs really fast"

Update: the last 2 days have been amazing, and the support ...

2026-06-06T08:04:03Z

Update: the last 2 days have been *amazing*, and the support folks have shown me is genuinely moving. I had another little cry.

But (oh, here he goes), there are still some places left on the upcoming workshops, and even without the discount that ended last night, they're worth every penny!

(1/2)

One last word: as I confess in my blog post, I'm in a real ...

2026-06-05T07:12:22Z

In reply to nevent1q…a0y2
_________________________

One last word: as I confess in my blog post, I'm in a real hole. No point pretending otherwise.

If you register for any of these workshops, not only will I be delighted to see you, but I'll also be very grateful.

Will train for food!

It would be really lovely to finish this chapter of my career playing to some packed houses

Finally, July 7-9, my flagship 3-day Code Craft workshop - 17 ...

2026-06-05T07:11:49Z

In reply to nevent1q…ecf9
_________________________

Finally, July 7-9, my flagship 3-day Code Craft workshop - 17 years in the making - goes on its final voyage.

I appreciate this one will require your boss's approval. But it's HALF-PRICE until midnight tonight if you book online if you can expense it.

https://codemanship.co.uk/codecraft.html

June 30th at 18:45 GMT+1, Specification By Example. There's ...

2026-06-05T07:11:23Z

In reply to nevent1q…4j48
_________________________

June 30th at 18:45 GMT+1, Specification By Example. There's no coding in this one, so testers, PMs, UX folks etc are very welcome to join us.

Again, until midnight tonight, just £75 + VAT. Bargain!

https://www.tickettailor.com/events/codemanship/2162096

On June 16th at 18:45 GMT+1 and June 20th at 09:45 GMT+1, a ...

2026-06-05T07:10:55Z

In reply to nevent1q…u20n
_________________________

On June 16th at 18:45 GMT+1 and June 20th at 09:45 GMT+1, a Refactoring workshop that packs so much into 2 hours that your head will spin :-)

Until midnight tonight, just £75 + VAT. Save ~£30.

Easily the second most important skill for a developer.

https://www.tickettailor.com/events/codemanship/2230313

So it's out there that - with great sadness - I'm winding ...

2026-06-05T07:10:20Z

So it's out there that - with great sadness - I'm winding down my training business & looking for a new challenge

But it's not quite over yet. I have 3 *final* workshops for developers (and, in one case, non-devs, too) coming up - all on special offer 2day

https://codemanship.wordpress.com/2026/06/04/a-car-crash-in-slow-motion/

Well, here we are finally - the sound of one door closing, and ...

2026-06-04T14:28:52Z

Well, here we are finally - the sound of one door closing, and the end of a 17-year journey.

https://codemanship.wordpress.com/2026/06/04/a-car-crash-in-slow-motion/

As LLM usage prices rise, the CFO is going to enter the chat and ...

2026-05-27T11:15:44Z

As LLM usage prices rise, the CFO is going to enter the chat and ask a simple question - what value is the company getting from this?

And you're not going to have a good answer, because you didn't ask that question yourselves.

Pro tip for event organisers: if you're asking a speaker or ...

2026-05-17T12:57:37Z

Pro tip for event organisers: if you're asking a speaker or trainer to do it in exchange for "exposure", asking them to promote the event is a bit of a tell.

I keep seeing posts on The Interwebs talking about how young ...

2026-05-17T03:59:15Z

I keep seeing posts on The Interwebs talking about how young people need to learn "AI skills".

Do they mean linear algebra? Vector calculus?

Or do they mean the ability to tell when the chatbot got it wrong?

In which case, "AI skills" means a good understanding of what you're asking it to do.

More bluntly, the ability to do it yourself.

The key to being effective with "AI" is being effective without it.

It pisses me off when even AI skeptics like Gary Marcus keep ...

2026-05-03T06:10:08Z

It pisses me off when even AI skeptics like Gary Marcus keep repeating the industry talking point that "AI has clear value in coding".

The best available evidence shows that the value is far from clear.

In fact, the best available evidence shows that the median outcome is net-negative.

I've joked about the possibility of human experts ending up ...

2026-05-02T03:35:27Z

I've joked about the possibility of human experts ending up being the better *and* the cheaper option to LLMs, but I also kind of meant it.

In Victorian textile mills, children as young as ten were ...

2026-04-24T19:16:43Z

In Victorian textile mills, children as young as ten were employed for up to 12 hours a day as "piecers", mending broken threads in mechanised looms. It was a dangerous job, often costing fingers and entire limbs. But it was necessary because - guess what? - the machines weren't entirely reliable.

In 2022, I didn't need a Large Language Model. In 2026, I ...

2026-04-24T12:36:36Z

In reply to nevent1q…87pn
_________________________

In 2022, I didn't need a Large Language Model.

In 2026, I still don't need a Large Language Model.

I intend to keep it that way.

Recent hiring trends suggest that employers may be favouring ...

2026-04-24T06:22:47Z

Recent hiring trends suggest that employers may be favouring senior developers with significant pre-AI experience.

As I’ve been only half-joking for over a year, the devs who’ll be in highest demand in this age of AI will be the ones who don't need it.

https://codemanship.wordpress.com/2026/04/24/the-hottest-ai-skill-in-2026-coding-without-ai/

Hiya. If you're hosted by fosstodon.org, have you received a ...

2026-04-18T22:51:32Z

Hiya. If you're hosted by fosstodon.org, have you received a warning about mastodon.cloud being defederated soon?

"Does code quality matter if LLMs can regenerate software ...

2026-04-18T09:15:59Z

"Does code quality matter if LLMs can regenerate software from scratch?"

LLMs can't regenerate software from scratch. That's not a thing, nor is it ever likely to be.

But thanks for letting us know you're getting your information about LLMs from LinkedIn posts.

True story: 100% successful autonomous completion of ...

2026-04-11T05:09:50Z

True story: 100% *successful* autonomous completion of non-trivial coding tasks by "AI agents" isn't a thing. It doesn't exist. It's not even close to existing.

Treat any such claims like you would treat claims of people teleporting to the office.

In software development, the limits of team size are the limits ...

2026-04-06T07:18:37Z

In software development, the limits of team size are the limits of effective coordination - whether it's a team of humans or AI agents.

As teams grow, the coordination overheard grows non-linearly, until coordinating - or dealing with the downstream consequences of not coordinating - becomes most of the work, and forward momentum slows to a crawl.

Don't act surprised. We've known this for more than 50 years.

https://codemanship.wordpress.com/2026/04/06/the-mythical-agent-month/

CircleCI's analysis of 28 million CI workflows confirms the ...

2026-04-01T04:11:20Z

CircleCI's analysis of 28 million CI workflows confirms the same picture the DORA data shows. While feature branch activity's up significantly, the median impact on *main* (i.e. release) branch activity's net-negative 7%.

Only the top 5% of teams saw significant gains. The top 10% flatlined at 1%.

For the average team, AI slows them down overall.

Told ya!

https://www.linkedin.com/pulse/what-28-million-workflows-reveal-ai-codings-biggest-risk-circleci-j9syc/

I'm pretty satisfied at this point that the true extent of ...

2026-03-31T11:54:08Z

I'm pretty satisfied at this point that the true extent of "AI" use in software development has been massively exaggerated.

Sure, lots of devs are using LLMs. But there seems to be very little advanced use. It's mostly chat window stuff and occasional inline completion, to keep the boss happy.

It is, however, a massive distraction.

Part of the obstacle is that if we were to set the bar at ...

2026-03-22T07:58:04Z

In reply to nevent1q…dkxk
_________________________

Part of the obstacle is that if we were to set the bar at "basically competent", 90% of the profession would have to significantly up their game.

Ever since I read Steve McConnell's After The Goldrush, ...

2026-03-22T07:36:38Z

Ever since I read Steve McConnell's After The Goldrush, I've been listening to people tell me why software engineering should be exempt from the kind of controls that e.g. electrical engineering or medicine or plumbing are subject to.

The reason is almost always "Their work can do real harm".

Yeah? Look around you, dude.

New rule: if your for-profit company is "too big to ...

2026-03-20T07:39:15Z

New rule: if your for-profit company is "too big to fail", you get nationalised, not bailed out.

I keep seeing the claim that "AI" multiplies the ...

2026-03-20T05:36:29Z

I keep seeing the claim that "AI" multiplies the productivity of senior developers repeated as justification for FOMO.

The best available data clearly shows that, in terms of *outcomes* at the *team* level, that multiplier is - at best - about 1.1x.

And for the majority, it's more like 0.9x.

We know exactly why this is, too.

1. Bigger change batches hitting bottlenecks like testing, code review and integration

2. Coding was only a small part of the process anyway

* sigh *

Because they have essentially the same causes, the buggier a code ...

2026-02-24T08:48:02Z

Because they have essentially the same causes, the buggier a code base is, the longer it usually takes to change (because changes keep breaking it).

The most popular "AI" coding CLIs seem to be *very* buggy - especially Claude Code. Thousands of open issues.

It's also probably very insecure.

I've half-joked that, in the future, the most in-demand ...

2026-02-21T08:18:40Z

In reply to nevent1q…v65x
_________________________

I've half-joked that, in the future, the most in-demand software developers will be the ones who haven't been relying on this technology. (Especially when vendors decide it's finally time to make a profit).

I'm taking the joke label off. I think it's probably going to happen.

The decline in program comprehension and reasoning I've ...

2026-02-21T07:43:34Z

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.

Nostr event nevent1qqst32s6re2kr423ue755p0sz2ergg5ye9cxj4masks3s60upurgvwszyqmdx27fzgvvw46uu5z6nlppr9nq9mdskku4t3gh2t3r99qlx3n3kfl3a63

2026-02-13T18:13:27Z

You realise that LLMs are Brawndo?

Increasingly I'm seeing posts from developers saying ...

2026-02-09T13:04:06Z

Increasingly I'm seeing posts from developers saying something along the lines of "hit my token limit, waiting for it to reset", and I can't help wondering if employers will soon see "I can work just fine without an LLM" as a plus on a candidate's CV.

But how do I ask them?

2026-02-05T21:37:14Z

In reply to nevent1q…95al
_________________________

But how do I ask them?

"Software developers won't need to understand syntax. ...

2026-02-05T18:43:32Z

"Software developers won't need to understand syntax. Just semantics"

I have questions.

Is this the new autocomplete drive I've been hearing so much ...

2026-02-02T23:07:16Z

Is this the new autocomplete drive I've been hearing so much about?

So you read that <your favourite programming thought ...

2026-02-02T15:26:55Z

So you read that <your favourite programming thought leader> is vibe coding now, and you've taken that as proof that "AI" produces great code.

I'd probably take it as evidence that <your favourite programming thought leader> is susceptible to social proof like the rest of us.

I regularly see folks hailing the end of the need to read code. ...

2026-01-30T12:07:08Z

I regularly see folks hailing the end of the need to read code.

If I presented you with a bowl of candy and told you 1% had broken glass in them, what % would you check before giving them to your kids?

The need to understand code isn't going anywhere.

https://codemanship.wordpress.com/2026/01/29/writing-code-may-be-dead-not-really-but-reading-code-will-live-on/

Software developers won't be replaced by AI. They'll be ...

2026-01-29T12:58:35Z

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.

One thing I'm keen to do in Q1 this year is find out how ...

2026-01-22T14:08:43Z

One thing I'm keen to do in Q1 this year is find out how developers are *really* using "AI" on production code.

I've been keeping one eye on it as I watch devs working, but I'd like to put both eyes on it for a while.

I get a sense that the practice and the rhetoric are some way apart.

It's the "Clever Hans" effect: folks genuinely ...

2026-01-19T11:39:04Z

In reply to nevent1q…haqv
_________________________

It's the "Clever Hans" effect: folks genuinely don't seem to realise how often *they* actually wrote/edited the code to get things working.

"Code is now disposable, you can regenerate systems in ...

2026-01-19T11:35:36Z

"Code is now disposable, you can regenerate systems in minutes", he says as I watch him going round and round in circles trying to get Claude to fix a bug.

My Dad's been in building design and construction for over 50 ...

2026-01-18T01:01:47Z

My Dad's been in building design and construction for over 50 years. He earns about the same as me.

His home office has shelves and shelves full of books about building regulations.

If he does bad design or gives bad advice, the consequences for him can be quite severe.

More than FIFTY YEARS.

We're basically children in software development.

The concluding post in my AI-Ready Software Developer series ties ...

2026-01-05T11:32:04Z

The concluding post in my AI-Ready Software Developer series ties all the threads together.

Far from "changing the game", AI coding assistants have just added another layer of uncertainty to an already very uncertain process.

Same game, different dice.

https://codemanship.wordpress.com/2026/01/05/the-ai-ready-software-developer-conclusion-same-game-different-dice/

"Hi. I'm an AI engineer". You engineer AI? "No, I ...

2026-01-02T07:07:52Z

"Hi. I'm an AI engineer".

You engineer AI?

"No, I use AI for engineering."

Hi. I'm a compiler engineer.

Most teams don’t see the impact of "AI" code generation ...

2025-12-30T09:42:47Z

Most teams don’t see the impact of "AI" code generation on systemic outcomes because THEY'RE NOT LOOKING.

https://codemanship.wordpress.com/2025/12/30/the-ai-ready-software-developer-18-productivity-you-keep-using-that-word/

On today's episode of I Haven't Thought This Through: ...

2025-12-29T12:40:36Z

On today's episode of I Haven't Thought This Through:

"LLMs should be generating code optimised to be understood by LLMs, not by humans"

U-huh? And what are we gonna train them on?

Feels like we should have some Christmas music. It doesn't ...

2025-12-24T23:14:29Z

Feels like we should have some Christmas music. It doesn't get much more Christmassy than this.

https://www.youtube.com/watch?v=qORYO0atB6g

As 2025 draws to a close, my hopes "AI" was going to ...

2025-12-24T09:09:50Z

As 2025 draws to a close, my hopes "AI" was going to force more teams to address the real bottlenecks in development fade into the distance.

95% will just ship less reliable software, and take longer and spend more £ doing it. And users will be coerced into eating the costs.

Business as usual.

The director of photography on The Holiday was Dean Cundey, who ...

2025-12-16T20:29:42Z

The director of photography on The Holiday was Dean Cundey, who was also DP on The Fog, Escape To New York and The Thing.

And that's totally why I'm watching it.

Watch software developers. I mean really watch them. When they ...

2025-12-15T08:24:02Z

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.

That was a fork in the road in my career. In those six months, I ...

2025-11-30T13:58:16Z

In reply to nevent1q…7agf
_________________________

That was a fork in the road in my career. In those six months, I went from "we just need to get it out!" to "we just need to make sure it works and will be easy to change", and - it turns out - *that's* how you "get it out" sooner.

It's like a magic trick.

After 6 months, I didn't need persuading. If a tradeoff ...

2025-11-30T13:53:04Z

In reply to nevent1q…tvyf
_________________________

After 6 months, I didn't need persuading.

If a tradeoff exists between quality and time-to-market, it's short-lived - perhaps in the order of days from a blank slate - before other factors start to kick in.

That "experienced" software professionals still need persuading of this in 2025 is...

*sigh*

Deja vu. The first time I saw the real relationship between ...

2025-11-30T13:52:14Z

Deja vu.

The first time I saw the *real* relationship between software reliability, cost of change and lead times was 30 years ago.

And we only collected that data because *I* didn't believe it.

As test assurance went *up* and the code became more modular (not unrelated), bug counts and cost of change went *down*.

We went from struggling to put together release candidates once a quarter to producing them almost every weeknight.

If you asked me to characterise software development in its ...

2025-11-28T06:28:26Z

If you asked me to characterise software development in its essence, I'd say it's a process of reducing uncertainty.

Otherwise known as "learning".

https://codemanship.wordpress.com/2025/11/28/we-dont-out-deliver-the-competition-we-outlearn-them/

Some folks say that they decided to focus on code correctness ...

2025-11-25T16:24:00Z

Some folks say that they decided to focus on code correctness instead of maintainability.

It's the darndest thing, though. Code that's hard to change tends to be broken, too.

Nostr event nevent1qqsyd6fug49nhkr2gttpkjsdqveeq9fgsgjhu7r64edqa70pdusekzczyqmdx27fzgvvw46uu5z6nlppr9nq9mdskku4t3gh2t3r99qlx3n3k7exsaz

2025-11-24T19:26:02Z

The future of software development is software developers.

If I was an employer of software developers, I'd be hiring ...

2025-11-22T10:26:47Z

If I was an employer of software developers, I'd be hiring now while I can have my pick.

There. I've said it. You may now start throwing the furniture ...

2025-11-07T11:25:14Z

There. I've said it.

You may now start throwing the furniture around.

https://codemanship.wordpress.com/2025/11/07/is-software-the-ufology-of-engineering-disciplines/

Teams organised around technology stacks and technical ...

2025-11-05T12:13:54Z

Teams organised around technology stacks and technical disciplines have about as much chance of achieving a business goal as a sack full of ferrets has of changing a lightbulb.

Smart companies organise around business outcomes.

https://codemanship.wordpress.com/2025/11/04/productivity-you-keep-using-that-word/

The most depressing posts I see on social media are from folks ...

2025-10-26T14:52:44Z

The most depressing posts I see on social media are from folks bemoaning their inability to achieve the amazing results using "AI" that they're reading about from other people. Obviously, they must be doing it wrong.

Pssst. Can I let you in on a little secret?

THOSE PEOPLE ARE PRETENDING.

Not judging, genuinely curious. All anon, of course. Software ...

2025-10-21T05:49:38Z

Not judging, genuinely curious. All anon, of course.

Software devs using "AI": have you exaggerated the extent of your use of AI coding assistants under pressure from management to be seen to be using them more?

I've joked before that "AI" coding assistants are ...

2025-10-20T07:34:57Z

I've joked before that "AI" coding assistants are like weight loss pills that only work if you exercise and eat healthily.

Some folks might interpret that to mean I think LLMs boost the performance of already-high-performing teams.

But what I'm *really* saying is I believe it's a placebo.

And my position as we hurtle towards 2026 is that if LLMs (ideally smaller OSS community-built ones running on consumer hardware) are what it's gonna take for more teams to work smarter, then so be it.

Not in the least bit surprised to learn "AI-created" ...

2025-10-05T10:09:27Z

Not in the least bit surprised to learn "AI-created" "actress" Tilly Norwood is actually just a statistical approximation of a real person, singer Stella Hennen, who - gasp! - didn't give permission for them to use her likeness and is now threatening legal action. She wrote a song about it first, of course.

Here's hoping this makes Stella famous and ends Tilly's "career" and this creepy face-stealing business model.

One argument I sometimes hear against trunk-based development is ...

2025-09-30T12:20:33Z

One argument I sometimes hear against trunk-based development is that the code may be changing rapidly, so it's better if developers work in isolation on their own branches.

That doesn't make a lot of sense to me. Surely it's the opposite? The faster the code's changing, the more often we should be merging?

"Teams that care about quality will take the time to ...

2025-09-30T07:52:35Z

"Teams that care about quality will take the time to understand LLM-generated code before it makes it into the repo.

But some teams are checking in code nobody’s read, and that’s only been cursorily tested. And, evidently, there’s a lot of them."

https://codemanship.wordpress.com/2025/09/30/comprehension-debt-the-ticking-time-bomb-of-llm-generated-code/

"This relates directly to encapsulation. We want the effect ...

2025-09-25T07:07:13Z

"This relates directly to encapsulation. We want the effect of a function to be contained in the same place – in the same class. When it isn’t, we end up having to share internal details with whichever other class is handling that piece of business – otherwise known as 'coupling'."

https://codemanship.wordpress.com/2025/09/25/what-are-the-objects-in-object-oriented-programming/

OpenAI finally saying the quiet part out loud. LLMs are as ...

2025-09-21T09:53:33Z

OpenAI finally saying the quiet part out loud. LLMs are as reliable as they're gonna get.

"experts have found that the problem is getting worse as AI models get more capable."

How can models be getting "more capable" *and* less reliable? More capable of bullshitting?

https://futurism.com/openai-mistake-hallucinations

You don't need to tell us about the character Nicholas Cage ...

2025-09-20T20:19:45Z

You don't need to tell us about the character Nicholas Cage is playing. You just need to give us a Scoville measurement for his performance.

Seeing posts from managers complaining their devs have given up ...

2025-09-18T08:57:19Z

Seeing posts from managers complaining their devs have given up on "A.I." coding assistants. "They must be using it wrong. Do they need training?"

At some point, we may have to admit there's no way to get a tune out of a ham sandwich. And it's not because we haven't had ham sandwich training.

Having a fun conversation with a computer science grad who ...

2025-09-11T09:51:44Z

Having a fun conversation with a computer science grad who doesn't need refactoring because they taught him C and C++ and how to access memory blocks.

You may have heard that LLMs are really good at programming. This ...

2025-09-11T08:01:57Z

You may have heard that LLMs are really good at programming. This is not true. They are bad at it.

But developers who routinely commit code that doesn't even compile, let alone that they tested, may think the code LLMs generate looks just fine.

And that, sadly, is a lot of developers. In case you were wondering how the myth got started about LLMs being really good at programming.

"Vibe Code Cleanup Specialist" is being added to a lot of ...

2025-09-10T13:17:58Z

"Vibe Code Cleanup Specialist" is being added to a lot of LinkedIn profiles. Looking at the bios, rarely see any mention of refactoring, legacy code, tests or separation of concerns.

I suspect "Cleaned-Up Vibe Code Cleanup Specialist" is coming next.

My advice? If there's one book you need to read to learn how ...

2025-09-01T14:29:16Z

In reply to nevent1q…remh
_________________________

My advice? If there's one book you need to read to learn how to rescue vibed code, it's this one. There's no magical "A.I." or "agentic" fix. Sorry.

https://www.oreilly.com/library/view/working-effectively-with/0131177052/

Seeing a lot of e-books appearing about fixing vibed code. One ...

2025-09-01T14:20:04Z

Seeing a lot of e-books appearing about fixing vibed code. One author has published 4 since the term was coined in February. Indeed, he's published 30 books since last summer, which he all totally wrote himself.

You might also pause to consider that when you need a ...

2025-08-31T06:08:01Z

In reply to nevent1q…sjwh
_________________________

You might also pause to consider that *when* you need a developer's help, they might not be so keen on cleaning up your vibe-coded mess as a career move.

And, since employers haven't been hiring and training new developers for 3 years, that skillset could be about to get much more expensive.

"Hoorah! We don't need those pesky software developers ...

2025-08-31T05:45:53Z

"Hoorah! We don't need those pesky software developers anymore". Really?

Pause to consider how you'll fare in competition against people who actually know what they're doing.

Reliable autonomous agents built on LLMs are never going to ...

2025-08-29T05:45:19Z

In reply to nevent1q…kech
_________________________

Reliable autonomous agents built on LLMs are never going to happen.

LLMs are a dead end. Not reliable enough for most use cases, and ...

2025-08-29T05:43:58Z

In reply to nevent1q…gymz
_________________________

LLMs are a dead end. Not reliable enough for most use cases, and never likely to be much more reliable.

A human with expertise will always be required to instruct and monitor the work. And, as we're seeing, the productivity gains just seem to evaporate when they're actually measured because these models tend to create more downstream work than they save us.

Why do I believe LLMs are about as reliable as they're going ...

2025-08-29T05:40:14Z

Why do I believe LLMs are about as reliable as they're going to get? Physics, my dear chums. Physics.

"For the case of α=0.1, reducing the error by a factor of 10 requires the allocation of ten billion more compute resources. And, if we focus on the power consumption itself, the exponent is 0.05, meaning an astronomical 10^20 more computing power for just one order better in accuracy."

https://arxiv.org/html/2507.19703v1

This is an exciting time in software development, as it slowly ...

2025-08-23T00:26:02Z

This is an exciting time in software development, as it slowly dawns on the money that they still need the talent, and the talent sees the money for what it was all along.

If you want evidence that using LLMs reduces your cognitive ...

2025-08-21T05:36:53Z

If you want evidence that using LLMs reduces your cognitive abilities, I offer folks claiming that "A.I." means they will only need senior developers going forward.

They seem to have become incapable of even the most basic causal reasoning.

The best summary of "prompt engineering" I've seen so ...

2025-08-19T17:40:32Z

The best summary of "prompt engineering" I've seen so far.

"We couldn't think of a way to add value, so we took ...

2025-08-17T20:20:37Z

In reply to nevent1q…4s0v
_________________________

"We couldn't think of a way to add value, so we took value away and you'll have to pay extra to get it back."

We live in an age where "premium" means "Not ...

2025-08-17T20:12:40Z

We live in an age where "premium" means "Not deliberately shit". What a time to be alive.

When I see posts on social media complaining about the ...

2025-08-17T08:32:18Z

When I see posts on social media complaining about the reliability of LLMs in generating code, there are 2 distinct kinds of replies from the "A.I." boosters that amount to:

* "You must be throwing the dice wrong."

* "Are you using the latest dice?"

They are never accompanied by a demonstration of the "right way".

You would hope that software developer job interviews and ...

2025-08-13T07:30:33Z

You would *hope* that software developer job interviews and software development wouldn't require two distinctly different sets of skills.

The solution to poverty is money. The solution to hunger is food. ...

2025-08-10T19:37:57Z

The solution to poverty is money.

The solution to hunger is food.

The solution to homelessness is homes.

The solution to illness is healthcare.

The solution to conflict is dialogue.

We don't need a fucking chatbot to solve these problems for us.

Here's how I see it: when we optimise a non-bottleneck in a ...

2025-08-09T09:26:26Z

Here's how I see it: when we optimise a non-bottleneck in a system with real bottlenecks, we make overall system performance worse.

Coding isn't the bottleneck in software development, but there are real bottlenecks in many dev teams' processes.

When they turn on the "A.I." coding firehose, they can overload their plumbing. Team productivity - measured in customer outcomes - gets *worse*.

=> The first step on your journey to "A.I."-assisted coding is to FIX YOUR REAL BOTTLENECKS.