Weeknote 61
Back to school officially. Distributed inference across AI or doing AI differently. Some really nice running, organisational resilience concepts. The programme really does work. Discussing bids, paying attention to what matters.
This week is officially back to school for my daughter but, also for me. I decided a little while ago to put myself back into an academic setting. It's been the first time in over a decade I've done any official learning apart from learning on the job.
One of the reasons for doing this was to give myself a bit of dedicated focus time to explore ideas in a bit more depth. What better way to focus than to spend an exorbitant amount of money on something? Money really does focus the mind on things, as you'll well know if you ever run free events versus paid events. The turn-up rate for people who pay for something is much higher. Let that be a lesson to us all.
I'm officially underway on that. It's going to be an interesting journey, having last done anything at this sort of level, probably a decade ago when I did things on strategic leadership and management or something like that. Me as a manager. Can you imagine? Some will say that traditional education is dead, and that whatever you are taught will have changed by the time is received. But I'm not doing this to be taught, I'm doing this to learn.
It's one of the reasons why I've been thinking so much about my attention at the moment. Where do I put my attention? I think over the past year I probably focused too much on trying to do things that shifted a wider system but most people in that system either aren't paying attention or just don't want to pay attention. That's why I let go of the grant-making stuff and that's why I'll probably be less active in a lot of the shifting of things. I'll still be putting stuff out there. I'll still be making things and doing things in my own way but I think I've kind of lost a bit of hope in that way.
I wrote some things over the last year, which I wasn't quite sure where they were coming from. One of them was on a rather esoteric journey on the cooling of a sector.
Images from 'The cooling of a sector' I did earlier in the year
I think that was me just really seeing things close up, things being abandoned, and that made me sad, but it's okay. I've got a new focus now.
One of them is my own studies. Another one of them is getting back into doing, paying attention, and focusing on the work that I enjoy with organisations that get it and take it on. This week I delivered a bit more of a session on organisational resilience. It was interesting. It's the first time I've delivered anything with a new framework since the programme ended in June and it worked. I knew it worked. I've tested this. I've done it with lots of organisations but it was nice to just do it in a single session and come back to it, having had a bit of a break from it.
And it just bloody works and that was nice to see. That was good.
Other things I did this week: I was meant to deliver a free session with Doug on TechFreedom. For personal reasons, I couldn't do that but Doug stepped in and handled that and delivered it. Thanks Doug.
TechFreedom stuff, I think, has always been a little bit on the edge of what a lot of people are comfortable with. I think one of the things we're considering is: where does this fit and sit? I think if you read anything that we've ever written about this, you'll see, in more and more of this idea of sovereignty and the risks involved in all of these things. But it's hardly at the top of most peoples todo list! It's hard and feels intangible. We've tried to make it tangible but where does it sit and is there actually a business model behind it? That's a tricky one.
Also discussed a couple of bids with people this week. I don't really go for tenders generally as I discussed below.
Tom Watson
Been working on the next or the first field station experiment:
How might we run AI a little bit differently?
This one is about shared inference. What does that mean? That means trying to run open models distributed across multiple machines. Open models are great, but you still need to have some hardware to run them and to run some of the bigger models, the more capable models, you need quite a lot of compute. Most hobbyists and most organisations don't have that anymore. We moved everything into the cloud so how do you do that?
We're going to allow people who join up to come together and use all our machines and try to run larger models across multiple machines over the web. What an interesting experiment! Yes it will not be as fast as a cloud model. It won't be as fast as a model run locally on your own machine but does it offer us an opportunity to think a little bit differently about how we work with AI? If everybody's going to use this can we try and think about different ways of doing that? That's what the experiments are about. That's a first in that we may extend it into other areas. We've got some other ideas around that but for now we'll see.
Here's a link. Come along.
I've been doing a bit more building of my own things. Been working on something called Attention, which is an agent-based system for "What should I be paying attention to?" I'm exploring multiple surfaces in which my attention can be directed. This is one of the things I'm especially interested in at the moment. This one agent now runs across for specific purposes: an adapted Rabbit R1, a web app, a Raspberry Pi, and a display on my own Remarkable. That's been fun.
Running update
This year hasn't been great for running. I started off slow at the start of the year, having kind of abandoned running at the back end of last year. I ramped up and was feeling okay before getting ill in the middle of April time. In fact that knocked off, apparently according to Strava, a 3.5-year streak of running at least once every week. That illness knocked me back a bit. I've been kind of struggling to really find my way with all of that in among all of my other kind of struggles. I think the past year I've lost myself in many ways, if I'm honest. I feel okay about that. I feel like I'm coming out of whatever that was and that's kind of reflected in my running.
A month. It's amazing what a month in the mountains will do for you. I just went out every day and ran and did whatever I wanted to do, and so to come back I'd not really notice any particular improvement in fitness until I got back onto my home trails. It's hard to gauge your fitness when you're running up an alpine peak in Austria. You never run up that peak. It's just hard, fun but hard, but then you get back on the trails.
This week I had two runs in the forest on trails I know and remember well. I felt good and strong on them, not maybe as strong as I once was, but as someone who's getting older, whose Garmin keeps telling them they're shit, it felt good. I felt strong and it felt enjoyable. I'm going to hold on to that.
Picked up my guitar again this week. It's been years. I think I'll do more of that. Pay attention to the things that matter.
Interesting things
- Damaged Earth Catalog
- Pride in Place practice - Lessons shared between Pride in Place areas: five dimensions of ambition for using £20m to grow something bigger, a self-assessment tool, and practical guides for chairing, consultation, storytelling and working with politics.
- Governance Analysis #2: How is Digital Sovereignty Measured Beyond Physical Location? - Ontario’s Data Centre Playbook treats digital sovereignty as a strategic pillar. But if sovereignty is only assessed by where servers sit, we risk confusing residency with control.
- From crisis to catalyst - Collective visions for civil society and philanthropy in 2026 and beyond
- Product for the People 2026 lightning talk: From A to Z — notes
- Care 2027: A Good Life in Every Neighbourhood
- Destination Earth - Building a highly accurate digital twin of the Earth
- Analogious joy - analogue versions of my Substack/Podcast - THE AMAZING EMPORIUM OF EVERYTHING - podcasts on cassettes?!