Jason Gorman

Live training workshop: Code Craft (3 days) - July 7-9

Strengthen your software engineering foundations to accelerate release cycles, shrink lead times and improve reliability - with and without AI

https://www.codemanship.co.uk

Business: https://www.codemanship.co.uk
GitHub: https://www.github.com/jasongorman
Bluesky: https://bsky.app/profile/jasongorman.bsky.social
LinkedIn: https://www.linkedin.com/in/jasongorman/

Public Key

npub1xmfjhjgjrrr4wh89qk5lcggevcpwmv94h92u296jugefg8e5vuds9w7m86

NIP-05 Address

jasongorman_at_mastodon.cloud@momostr.pink

Profile Code

nprofile1qqsrd5eteyfp336htnjst20uyyvkvqhdkz6mj4w9zafwyv55ru6xwxcprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmntqtwl2g

Publishing to

relay.momostr.pink

Author Public Key

npub1xmfjhjgjrrr4wh89qk5lcggevcpwmv94h92u296jugefg8e5vuds9w7m86

Show more details

Last Notes

2026-03-22 07:58:04 UTC

- reply

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.

2026-03-22 07:36:38 UTC

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.

2026-03-20 07:39:15 UTC

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

2026-03-20 05:36:29 UTC

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 *

2026-02-24 08:48:02 UTC

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.

2026-02-21 08:18:40 UTC

- reply

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.

2026-02-21 07:43:34 UTC

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.

2026-02-13 18:13:27 UTC

You realise that LLMs are Brawndo?

2026-02-09 13:04:06 UTC

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.

2026-02-05 21:37:14 UTC

- reply

But how do I ask them?

2026-02-05 18:43:32 UTC

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

2026-02-02 23:07:16 UTC

Is this the new autocomplete drive I've been hearing so much about?
https://media.mastodon.cloud/media_attachments/files/116/003/544/157/988/213/original/63bbf377f716f489.png

2026-02-02 15:26:55 UTC

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.

2026-01-30 12:07:08 UTC

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/

2026-01-29 12:58:35 UTC

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.

2026-01-22 14:08:43 UTC

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.

2026-01-19 11:39:04 UTC

- reply

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.

2026-01-19 11:35:36 UTC

"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.

2026-01-18 01:01:47 UTC

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.

2026-01-05 11:32:04 UTC

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/

2026-01-02 07:07:52 UTC

"Hi. I'm an AI engineer".
You engineer AI?
"No, I use AI for engineering."
Hi. I'm a compiler engineer.

2025-12-30 09:42:47 UTC

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/

2025-12-29 12:40:36 UTC

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?

2025-12-24 23:14:29 UTC

Feels like we should have some Christmas music. It doesn't get much more Christmassy than this.
https://www.youtube.com/watch?v=qORYO0atB6g

2025-12-24 09:09:50 UTC

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.

2025-12-16 20:29:42 UTC

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.

2025-12-15 08:24:02 UTC

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.

2025-11-30 13:58:16 UTC

- reply

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.

2025-11-30 13:53:04 UTC

- reply

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*

2025-11-30 13:52:14 UTC

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.

2025-11-28 06:28:26 UTC

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/

2025-11-25 16:24:00 UTC

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.

2025-11-24 19:26:02 UTC

The future of software development is software developers.

2025-11-22 10:26:47 UTC

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

2025-11-07 11:25:14 UTC

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/

2025-11-05 12:13:54 UTC

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/

2025-10-26 14:52:44 UTC

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.

2025-10-21 05:49:38 UTC

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?

2025-10-20 07:34:57 UTC

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.

2025-10-05 10:09:27 UTC

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.

2025-09-30 12:20:33 UTC

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?

2025-09-30 07:52:35 UTC

"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/

2025-09-25 07:07:13 UTC

"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/

2025-09-21 09:53:33 UTC

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

2025-09-20 20:19:45 UTC

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.

2025-09-18 08:57:19 UTC

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.

2025-09-11 09:51:44 UTC

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.

2025-09-11 08:01:57 UTC

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.

2025-09-10 13:17:58 UTC

"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.

2025-09-01 14:29:16 UTC

- reply

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/

2025-09-01 14:20:04 UTC

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.

2025-08-31 06:08:01 UTC

- reply

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.

2025-08-31 05:45:53 UTC

"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.

2025-08-29 05:45:19 UTC

- reply

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

2025-08-29 05:43:58 UTC

- reply

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.

2025-08-29 05:40:14 UTC

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

2025-08-23 00:26:02 UTC

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.

2025-08-21 05:36:53 UTC

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.

2025-08-19 17:40:32 UTC

The best summary of "prompt engineering" I've seen so far.
https://media.mastodon.cloud/media_attachments/files/115/056/653/365/126/057/original/e9fa9745df452463.png

2025-08-17 20:20:37 UTC

- reply

"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."

2025-08-17 20:12:40 UTC

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

2025-08-17 08:32:18 UTC

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".

2025-08-13 07:30:33 UTC

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

2025-08-10 19:37:57 UTC

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.

2025-08-09 09:26:26 UTC

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.

2025-08-08 05:26:22 UTC

"I speculated that transformer performance would converge on not-quite-good-enough. Needs more work. See me after. Not so much 'super-intelligence' as 'super-mediocrity'."
#gpt5
https://codemanship.wordpress.com/2025/01/11/the-llm-in-the-room/
The time is upon as, folks. If anyone doubted that LLMs have hit a performance wall, it's undeniable today. This is as good as they're gonna get, and it ain't good enough.

2025-08-06 06:20:07 UTC

Two principles of code readability I had drummed into me in my mid-20s:
1. "Comprehending" code means being able to correctly predict what it will do without running it.
2. A unit of code is "comprehensible" if you can predict what it will do without having to look inside any other units it uses. Boxes should be meaningfully labeled so you don't have to open them to see what they do.

2025-08-03 08:14:23 UTC

- reply

Perhaps the built-outs are the whole point of it? Picks and shovels etc.

2025-08-01 12:04:40 UTC

It's a classic failure mode when your tests all look like this - big set-ups, lots of mocked dependencies.
And it's understandable when folks interpret this is a test code problem. And, to be sure, if your tests look like this, YOU HAVE A PROBLEM. They bake in the internal design, making change much harder.
But the underlying cause is the real culprit here: bad modular design. The temptation is to rely on some fancy mocking to get around the problem. But that makes it worse.
Go to the cause!
https://media.mastodon.cloud/media_attachments/files/114/953/397/672/481/961/original/c85d1f1724247d00.png

2025-07-31 08:16:54 UTC

- reply

In case you were wondering why almost everybody's lying about the amazing things they're achieving with these tools when they think the boss might be listening (or the recruiter might be checking their LinkedIn page).

2025-07-31 07:28:35 UTC

The C-suite is mandating engineers use "A.I.", and demanding the productivity gains and cost reductions the vendors and consultants promised them.
Meanwhile, prod is on fire 24/7, engineers are abandoning ship left, right and centre, and all other work has been shelved.
LOL.

2025-07-31 05:40:19 UTC

ChatGPT has basically been turned into this guy
https://www.youtube.com/watch?v=KRPcssq-7Us

2025-07-27 16:21:23 UTC

"Dear Dr. Gorman, Jason,
Based on your outstanding contributions and significant achievements in your field, we are honored to extend a personal invitation to you for the Global Excellent Scientists Fund 2025 (Supplementary Round)."
Seems legit.

2025-07-22 08:17:40 UTC

I still get a real sense, from watching developers work (and bearing in mind that the act of me just watching probably changes their behaviour - though not as much as I'd hope), that there is a wide gap between what we say we do and what we actually do.
We get glimpses, like that large-scale study of testing activity in the IDE that found that maybe 8% of devs who *said* they did Test-Driven Development even did anything close, and I know that "refactoring" very rarely means actual refactoring.

2025-07-21 12:11:50 UTC

Maybe should have asked ChatGPT about "the tortoise and the hare"
https://media.mastodon.cloud/media_attachments/files/114/891/145/388/161/190/original/29e3dc33656b754b.png

2025-07-21 04:10:32 UTC

https://media.mastodon.cloud/media_attachments/files/114/889/256/264/303/397/original/3ef4cde6b3e3da6a.png

2025-07-16 11:16:12 UTC

When I openly express doubts about the benefits of "A.I."-assisted software engineering, I will often get responses that really do sound a heck of a lot like "Shut up and drink the Kool-Aid!"
It's looking increasingly like a religion.

2025-07-14 11:56:07 UTC

One good thing I got from doing a degree in physics is an inoculation against magical thinking.

2025-07-09 14:57:22 UTC

Anyone else been hearing rumours of... I guess we could call them "ghost coders"?
These are developers brought in by companies who publicly megaphoned they're going "AI-first" and have been quietly sneaking contractors in by the backdoor to fix things, at apparently very high rates.
I speculated that this might happen, but is it already? Or urban myth?

2025-07-04 08:07:30 UTC

Anyone else hearing from devs that they're being "strongly encouraged" to use LLMs by management - who are also raising expectations about "productivity gains" and "cost savings" - and they're not at all happy about it?
Is "A.I." just the latest stick to beat teams with?

2025-07-02 07:40:59 UTC

"LLMs will get better" - they got worse
"LLMs will get cheaper" - price plans are rising *fast*, and they're just getting started
When I list the Pros and Cons of the technology, most of the Pros are hypothetical future developments. All of the Cons are already happening.
This could well be as good and as cheap as they'll ever be.
#nocigar

2025-07-02 05:20:32 UTC

LinkedIn, it turns out, was invented so we could find out what realtors - sorry, "AI experts" - think about the future of software development.

2025-06-29 18:56:05 UTC

Shout out to all the European pioneers in computing and software engineering, without which Silicon Valley would just be... well... Valley.

2025-06-26 08:00:47 UTC

"Even if OpenAI, Anthropic etc stopped training new models, we'd still have access to the ones we have now".
Okay:
1. You gonna host them?
2. Enjoy working in Python 3.10 *forever*!

2024-11-12 09:04:06 UTC

"TDD is slow because you're writing twice as much code"
Okay, let's pick this apart.
First of all, how much of our time is actually spent writing code? It's typically a lot less than devs think.
Secondly, how much of our time do we spend debugging code? It's typically a lot more than devs think.
Thirdly, would it be less time writing code if we wrote the tests after? Can't see how that follows.
So, it's not TDD, is it? It's writing automated tests you have the problem with.

2024-11-10 09:55:00 UTC

Lazyweb: I know that, technically, any computer program can be represented as a finite state machine, and we can model state machines if we choose to, but is there a programming language or syntax that explicitly embodies that paradigm?

2024-11-10 09:12:58 UTC

Fun fact: everything Elmo has done to reinstall That Man in the Whitehouse was enabled by software developers.
Sleep well.

2024-11-09 22:43:23 UTC

So the VP-elect is threatening to withdraw the US from NATO if the EU doesn't bend over for Elmo's business interests in Europe?
Didn't take long for them to show their hand.

2024-11-09 07:36:52 UTC

Many of the software start-ups I've engaged with over the years didn't have anything even approaching a business plan. If they'd been opening, say, a bakery, no investor or lender would have looked at them twice. Because real businesses have to have a plan to make a profit.

2024-11-07 20:00:29 UTC

Silver lining: Elmo is about to find out what it's like to work for him

2024-11-07 19:46:52 UTC

My people are software professionals, sci-fi fans, cinephiles, progressive metal peeps, science folks, engineering stans, writers, actors, painters, architects, historians, philosophers and dreamers.
Wanna hear a "great replacement theory"? We're being replaced by dumbasses.

2024-11-07 14:33:33 UTC

It occurs to me that Cloud services are a major revenue stream for Silicon Valley. But I'm wondering just how many companies outside the US are going to continue to use them?

2024-11-07 04:53:12 UTC

Is it just me, or does the level of entitlement among software peeps grow year-on-year?
"What are the principles of software development? Teach them to me now!!!"

2024-11-05 22:57:30 UTC

It's interesting how quickly Musk stans will resort to abuse.

2024-11-03 11:06:00 UTC

- reply

A bad programmer will write bad code in any paradigm 🙂

2024-11-03 10:51:57 UTC

Fun fact: every single time a programmer told me they abandoned OOP in favour of FP because "OO code is bad!", when I looked at their FP code, that was bad, too.
Wherever you go, there you are.

2024-10-28 03:51:24 UTC

One criticism I hear often about object-oriented programming is that it tightly couples functions to the data they use.
I'm still waiting to see an example of a function that isn't coupled to the data it uses.
Does it not count when the data is parameters? Because that's *double* coupling: now the client knows about the data the function's using, too.

2024-10-27 14:42:48 UTC

- reply

It's by far the most impactful factor in our chances of success, and should therefore be right of the top of our list of questions before we consider proceeding.
With engaged users, ready to give feedback on working software as often as needed, the rest is up to us.

2024-10-27 14:42:41 UTC

When I think of the foundational practices of software dev:
* Iterative & Incremental Delivery
* Usage-driven Design
* Continuous Testing
* Continuous Integration
* Separation of Concerns
There's only one factor beyond our control: how often can we get feedback from real users?

2024-10-26 07:19:54 UTC

A classic anti-pattern I've observed many times (and endured more than a few) is when you hire a dev team who have a strong reputation for software delivery, insist they do it *your* way, and then wonder why you're not getting the same results.