Jon Ayre

Leaders. There's more than one way to scale an organisation, and how you do it depends on what you're scaling for. Are you scaling for brains or for brawn? Either can make sense but you need to decide, because how you scale will dictate what you get.

Apparently, folding extremely expensive and delicate display technology in half on a regular basis is not a great idea. Who knew?

When criticising GenAI writing there's a tendency to focus on sentence structure and punctuation, but neither is the issue. LLMs learned these structures from humans, and there's nothing particularly "AI" about them when used properly. It's just that AI doesn't do that.

Ondřej turned a "404 - love not found" error into a flood of "200 - deep appreciation returned" responses

The new Copilot advert is a great lesson in why AI won't deliver the promsied productivity gains. If you haven't seen it, it shows two stereotypically competitive guys in an office vying to create the most beautiful report. One is ahead, but the other tells copilot to "make it beautiful". He wins.

The software industry doesn't need to "reinvent software engineering". It just needs to adopt the practices of the engineering part, rather than just borrow the name. Engineers understand the things they build, even when automation plays a heavy role in the process.

I woke up this morning with the memory of the 1940 animated classic Fantasia. Specifically, the section entitled "The Sorcerer's Apprentice". That sequence seems very relevant to current events in the software industry. If you haven't seen it, give it a watch. I think you'll see what I mean.

Never has "too good to be true" been more relevant

I see LinkedIn has added a new option for "liking" posts - it's called "recommend". What I think they should do is also offer an equivalent negative set. Do I think this would be a good feature? No. Do I think this would enhance the platform? Also, no. But the chaos would be fun while it lasted.

The lyrics from 225 by New Model Army. Released in 1989 - still relevant today. We need more bands with real lyrics about stuff that matters.

Probably best they're gone, given the nature of the banner image which is labelled as a Character.AI screenshot. They should probably check the harddrive of the guy who wrote the prompt for that one.

Will

Or, and hear me out here, children who live in a stable and happy environment are more likely to have the time and inclination to read. Positioning reading as a solution for children in miserable environments might be missing the real issue.

Heading to London for Fast Flow Conference. Maybe see you there if you're going. You'll be able to spot me - I'm the one who looks like the father of the guy in my profile picture 😉

This is what we engineers call a single point of failure.

Andy Burnham confirms what previous PM's - May, Johnson, Truss, Sunak and Starmer - failed to, that Russia interfered in the Brexit referendum.

I took a look at what AI has to say about my continuous evolution approach to strategy. As expected, it gives a decent summary of the approach, including links to the source material. It then asks if I'd like to drill into specific detail, so I chose the Roadmap Radar. It goes downhill from there.

Anthropic's rule about abusing its language models has nothing to do with "cruelty" or "behaviour" & everything to do with reducing the amount of filtering they have to do on the training material before they use it That training material being your chat histories Take a deeper look at those T&Cs

"Preparing for" or "preparing to"?

Harvard just dropped a study AI productivity in software engineering and it found the average increase in productivity from using Claude code is.....zero and for exactly the reasons you'd think fion.ac/jellyfish.pdf

This is why we shouldn't let Americans build technology.

Did you know you only need a small brain to make silly little drawings?

The short version: AI changes how the job is done, but not who does the job.

Remember this if you're spending your company's money on AI. You're spending it in the hope that it will deliver productivity gains in the future, not now. It's an investment, or more accurately a bet. The bet might pay off but the evidence isn't there yet to suggest it will.

Less an iron dome, more a tin sieve.

Haha! This is the very finest satirical take on ridiculous TV game show formats... ...oh!

Exactly this. If you know anything about LLMs, you know that the media's reporting of it is hopelessly inaccurate and ignorant of the facts. Apply this awareness to reporting on all fields, including the ones you don't know anything about. It's all just clickbait.

Don't worry. You're not going to get left behind. Don't let the FOMO peddlers get you down.

This has real "The Office - You're still thinking about the bad news" vibes about it.

Four attempts to run an unchanged GitHub deployment pipeline that has always worked fine before. Different failure point each time. Changed nothing & ran the workflow again. On the fourth attempt it worked. My (in)sincerest thanks to the GitHub team for embracing AI assisted development!

If you roll a dice and every time it comes up 6 you do something stupid, it's not the dice that needs regulating. #JustSaying (Yes - I know dice is plural and die is singular but using die just looks pretentious 😉).

Hmm... jonayre.uk/blog/2022/12...

There's an "if it's too good to be true" aspect to this. People have become accustomed to getting significant output from models for very little financial outlay. That feels too good to be true, and as the saying goes, that's because in the long run it almost certainly is.

Not convinced about Man City's new kit for this season tbh