It’s been 29 weeks straight of posting these weekly things. I’m amazed that I’ve kept at it. It’s been kind of helpful to formalize the things I’m thinking about each week, but I also know that I’m not digging deeply into things. Quantity over quality, as it were. Although, I don’t know that it’s these weekly things that are getting in the way of that - are they like some kind of textual relief valve? Hopefully not. Anyway. I’ll probably keep this up for a bit more - it’s been super useful to have links searchable at least. But I could do that offline in Obsidian or something. 🤷‍♂️

I’ve used a custom, hand-rolled homepage ever since Google googled its iGoogle personalized homepage thing back in 2013. It’s set as the homepage for my browsers, and is a quick way to get to things that I use often and to check in on weather and status at several places at once.

Screenshot of the original version of my homepage.

But it never quite worked the way I wanted it to. It wasn’t super responsive on mobile devices, not scaling or adapting the layout very effectively. It used cobbled DIY html and CSS, and I’m sure it could be improved.

I am rather full of self-loathing after spending so much time using generative AI tools today. But this was actually really useful.

Our team uses a List in our MS Teams site to track projects - their status, contributors, due dates, etc. It started as a small list that was pretty manageable, but has grown into a list with 99 items. Data has been entered organically, so it’s not like it’s a fully normalized database with rigid taxonomies etc. So, making sense of it has become more of a challenge.

I followed this howto on Innovation for Bytes¹ to set up Ollama and connect it to the Smart Connections plugin. It had the missing bits of documentation for the Model Platform and path bits.

After Ethan Mollick’s keynote at SCUP 2024, I decided that I need to spend much more quality time with generative AI tools so that I can better understand how they work and what we can (and can’t) do with them. I still firmly believe that running LLM tools locally is important - if I’m feeding an LLM my content (some of which can’t be shared with third parties just due the nature of my work), I need to know it’s not being shared. The only way to ensure that is by running everything locally. Which means lots of trade-offs.