Thought Shrapnel

Tom Watson, who apart from doing generally awesome stuff somehow also has time to star in a documentary about ultrarunning, saw a recent Thought Shrapnel post and wrote about what tech he’s using.

I need to do my own, and actually Tom’s post has made me realise the extent to which I’m dependent upon Google and, to a recent lesser extent, Perplexity.

Prompted by Doug… and a couple of the Colophon’s (new word for me!) by Matt and Steve I thought I’d outline “my stack”. I’ve done this a couple of times before but this time feels slightly different.

[A]s someone who has been gently prompting people to not be so beholden to Big Tech, to look more at Open Source, to think more ethically, and to at least consider European Alternatives, I feel I should at least discuss where I’m trying to do this, where I’m succeeding and where I’m often failing.

[…]

…I use AI for specific things when I think they will make something more effective. I’m therefore always looking for the best model, and best use case. And things change all the time. So I purposefully build specific components that allow me to easily switch models and providers. If there is one thing I’d advise when thinking about building AI into an organisation, it’s to ensure you aren’t creating provider lock in for yourself. Quite a few AI wrappers (tools that put some kind of front end onto a model) allow you to switch models. But not all. And if you are building yourself, there is a risk you just lock yourself into a depeciating model or a provider that just turns out to be mega shitty.

[…]

It’s not easy to avoid the big tech trap, but I think I’m doing ok. Also I’m not saying you definitely should, but I think you should at least consider what you use and what this means, and if you have principle maybe they should cost you something.

Source: Tomcw.xyz

Image: Matthias Reumann