Sign in Subscribe
By Tom Watson in weeknote —

Weeknote 47

Another shorter week, trip to northern ireland, delivering a workshop on question based approaches to data, meeting new people, prepping for TechFreedom first cohort, making Bearing a tool for choosing AI models.

Weeknote 47
The Giants Causeway - looking down on the stones.

So it's still the Easter holidays up here, so a shorter week, only working 3 days. Feel like I've crammed a lot in, with a quick visit to Northern Ireland at the end of the week which was lovely.

What I did

  • Spent time planning out the Collective Resilience sessions as part of the LBF OR programme. We've spent a lot of time on the Organisational Resilience part of this programme and feel it's been both well designed and useful. But we know that no matter what organisations do themselves, some things are systemic, and that there needs to be a focus on Collective Resilience. This is a more difficult design, things are so very different in every place, and taking action is harder, because it's not just you and your organisation. But I'm very conscious that even though it is harder, we don't need to make it complicated. Honestly I think there is too much waffle in systems change stuff, too many consultants talking obscure language and fancy diagrams.
  • Delivered a workshop for Sheffield Futures as they embark on a programme of work to improve how they learn, how they think and talk about data. It was with a whole range of people across the organisation, many who had NO interest really in data. And that's precisly why I talk about curiosity and questions. The question based approach was pretty much the foundation of Data For Action, and just because DFA is going away, it doesn't mean the approach is. Kirsty talked about the work better than I ever can
#curiosityculture #charityslog | Kirsty Sandamas | 10 comments
Spending the last day of my 20s getting messy... with post-its, and questions, and challenges to explore! We’ve embarked on a curiosity journey at Sheffield Futures, exploring the richness of our data and the stories we have at our fingertips, as well as those stories we’re not currently telling but probably should be. Looking for the gaps, validating our lines of enquiry, and using metaphors to help shape our collective understanding of what data actually is and how we can use it to impact change. When you’re so busy doing the work - dropping that pebble in the lake - sometimes you only have the time or the energy to pay attention to the immediate ripples. But what of the wider, more subtle ripples that continue long after you’ve stopped looking? It’s in these waves that societal shifts can emerge. We can capture insights that shape how we work with young people everyday, and how we can influence change on a wider level. That’s wildly and wonderfully important. With huge gratitude to Tom Watson for being our guide on this journey 🧭 (and t’other Tom Tom French for lending the map) and to Prudence Trust for enabling us to embark. P.S. Last pic is my energy for tomorrow and the other journey I’m taking: entering my thirties. 😎 Out of office is on for the rest of the week as I wrap my head around this new concept. #CuriosityCulture #CharitySlog | 10 comments on LinkedIn
LinkedInKirsty Sandamas
The question based approach from Data For Action
  • You can run the above approach with post its, a canvas, a conversation or a free online tool https://www.questionsforaction.com/
  • Met Graham from the Movement Research Unit for a great chat about TechFreedom and a range of other things. I love meeting and learning about things I never knew existed (like the MRU) and seeing the really interesting things they do!
  • Spent some time finalising session 1 for the TechFreedom cohort. It starts next week and I'm a mix of excited and nervous. I think it will be good, but I'm certain there will be things we will learn from. You can still sign up for the cohort (or the newsletter if you're still on the fence)
  • Found out that we'll be running a TechFreedom session at TechNext in June
  • Spent a bit of time playing around with the Volunteering Data repo which was put together by ODI and others. I'd read a post about it the week before, so had a poke around, ending up building a working MCP server and then creating a demo dashboard of some of the data via an LLM. It was an interesting experience, seeing what was possible, and where there are gaps both in the data and in the infrastructure.
GitHub - dataforaction-tom/volunteering-mcp-server: MCP server for the Volunteering Open Data API — gives LLMs access to volunteering activities, organisations, and causes
MCP server for the Volunteering Open Data API — gives LLMs access to volunteering activities, organisations, and causes - dataforaction-tom/volunteering-mcp-server
GitHubdataforaction-tom
I read a really interesting post last week from Paul King talking about the volunteering data standard work done by the ODI and others. Paul talked about the good, the bad and the unfinished. Yes… | Tom Watson
I read a really interesting post last week from Paul King talking about the volunteering data standard work done by the ODI and others. Paul talked about the good, the bad and the unfinished. Yes there is a standard , some data even if lacking and an unfinished MCP server all in a github repo. Oh I do love a hacky bit of stuff to play around with... So I spent a couple of hours poking around in the repo, and ended up building a fully working MCP server just for kicks. And then had a play around with the data via claude. I asked about volunteering near me, what data was held about that in the database and what insights could we create from it, and while you’re there, give me a dashboard please mr claude (this wasn’t the prompt, but I think you know that) Anyway, here’s a dashboard I was able to spin up. Now, I don’t know much about the process around coming up with the standard, who was involved, etc. It’s really not my area of expertise (although having played with the data I have some thoughts). But was most interesting to me is that, well there IS a standard. And if we have a standard, and some api’s we can DO so much. It got me thinking about work I’ve done recently with Tech4Good South West around Digital Inclusion. And the airtable MCP server I built for that project. And how with some kind of standard, and a little bit of technology, we can do just in time interfaces whenever we want. Which is fun. Anyway, link to post by Paul in the comments.
LinkedInTom Watson

New tool

Built a new tool called Bearing. Bearing is an AI model recommendation tool which asks you to describe what you want to use AI for, choose what matters most to you (quality, speed, transparency etc), and Bearing will rank the best models for your specific task. I built it to hopefully challenge people to not just default to the most expensive or frontier models when they don't need to, and to perhaps embrace smaller and open source models, which cost less both financially and resource wise. There are hundreds of models out there, but most will only ever encounter one of the big three (claude, chatgpt, gemini). I think tech should be about informed choices, so this is one way I hope to bring a little light to knowledge around AI models.

In my own work, I'm constantly looking at which model is best for which job, often combining models in a tool. Even as someone doing this a lot, it's tricky to always keep up to date, so this was a tool and a thought exercise for me as much as anyone else. I found defining the scoring hard, and it's very much a work in progress.

Part of the other reason for building the tool is to build a bit of an open data set around what people actually want to use AI for. Whenever anyone uses the tool we create a record of the type of thing they want to use AI for, not the actual prompt, but an anonymised categorisation. I think this dataset could be useful, and is live and open on the site.

I've built the tool as transparently as possible, the scoring, the methodology, the code, all open. I hope others will help to improve it. You can see the tool below and read the documentation here and find the repo here.

Bearing — Find the right AI model
Transparent AI model recommendations. Describe your task, set your priorities, get a ranked shortlist scored across 7 factors.
Bearing

Has a visit to Northern Island at the back in of the week. Stunning scenery at Giants Causeway and the coast. Visited the Old Bushmills distillery and went extreme and had a 21 yr old bushmills which was incredible. Had at least one customary Guinness

Subscribe to Tomcw.xyz

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.
[email protected]
Subscribe