Maybe Terry

It makes me happy to know that you exist. Life would be a far grimmer place without your spark of light throwing back the darkness.

Try https://youtube.com/@w2aew for everything you could ever want to know about scopes.

Safe to assume that the majority of content on LinkedIn is now written by machine and therefore the value of the platform is rapidly descending to zero. The future is AI agents bragging fictitiously to other AI agents until someone eventually pulls the plug.

Another important reference is ‘The New Argonauts:
Regional Advantage in a Global Economy’ by AnnaLee Saxenian, 2006 Harvard University Press

Yes, opened as ‘The Institute’ in late 1990. Industrial decor in the main room with lots of broken mirrors. Andy Taylor was one of the investors. I did a lot of work with the previous club that the team had built in Halifax, called the Coliseum, so ended up helping out at the Institute on and off.

Does it still exist? I was there when it was converted to a club, back in the dim and distant when I was a lighting designer in the leisure industry. The brings back dusty memories of crawling around inside the organ.

Although Digbeth Institute would have been an ideal venue for that experiment since I seem to recall that being mostly UV lights and mirrors…

Seems like an ideal excuse to buy a cheap UV torch and host your own private disco.

You could try looking at yourself under UV light to see if you have striped or checkerboard skin.

To reach artificial intelligence, we must necessarily first pass through artificial stupidity.

The real question is whether we can survive the businesses who believe that they can make money out of deploying artificial stupidity at 1/10th the cost of real stupidity.

The delivery approach of an organisation is intrinsically linked to the budgeting and reporting strategy of the organisation. If you plan and set bonuses annually, you end up building product on an annual schedule and in a way that best games the bonus metrics.

And this, dear reader, is how the great orb infestation of 2025 began…

I have had to build resources like https://bestpractices.cd.foundation and https://bit.ly/MLOps2024 to even start to bridge the gap in understanding between one domain and the other.

The drivers between academia and industry are fundamentally misaligned. In the former you must demonstrate the validity of an algorithm and in the latter you are much more concerned about the quality, safety and maintainability of the asset being created.

Sadly this is an ongoing bug that has persisted for many years without resolution. It is made worse by having more than one of the same monitor attached, and by having monitors that are slow to wake.

You may be able to mitigate this with EDID emulator dongles or just move stuff back with scripting.

The 2024 edition of the MLOps Roadmap is now available. Check it out at https://bit.ly/MLOps2024

Hence the importance of having clearly defined outcomes and outputs in advance, so that you can detect the drift to output metrics and correct it before it becomes the dominant behaviour.

You have to make the difference explicit before you start or the hard thing won’t get done.

If you are measuring outputs, you can appear to be highly productive, yet actually be really busy doing the wrong things.

No none will pull you up on this until you have spent all the money and nothing has changed.

When you are running a change programme, you need to be measuring the effectiveness of your actions against your outcomes but in practice outcomes are hard to measure.

What inevitably happens is that people conflate outcomes with outputs and start measuring their outputs instead, since that’s easy.

There is a good reason from an implementation perspective. May I have your permission to contribute to this thread?

An object lesson in what happens if you don’t version your APIs but are contractually obliged to update them annually.

There should be a word for that complex moment when you discover that someone who is a long term part of your social media landscape is in your neighbourhood, but you don’t actually know them so can’t offer hospitality without it being weird. 😊

Training a 1 trillion parameter model is a 1GW exercise, but this is an infrequent process. Inferencing, however, is a continuous load that is expected to grow logarithmically based upon very high volumes of realtime data flowing from every edge device on the network.

Unfortunately, this is a massive underestimate. The real calculation needs to take into account the energy cost associated with flipping each bit in every transaction connected to the data centres.

See section 3.3 of https://irds.ieee.org/images/files/pdf/2023/2023IRDS_SA.pdf

Best used at lower temperatures with a hot air gun. Keep in mind that if you cook a board, it will make everything in the room smell absolutely terrible for months.

YouTube video by The Children's Literature Podcast I Talk to an Immigration Judge about Paddington's Passport

In a small piece of good news, the British government has issued Paddington Bear a passport so that he will not have to be deported to Rwanda. youtu.be/-MpxhQ6eJUY?...

From outside, the experience of those interactions between you and Wizard Cal and Lily feels like being party to a group of best friends who have been going on adventures together since they were children, and being able to pretend for a moment that you are yourself a part of that inner circle.

I remain eternally grateful for your ability to always make me smile with innocent joy. Never let anything change the way you see the world!

My immediate thought too. Your little corner of Twitter was the best space in the world.