Q2 2025 quarterly review

highlight
productivity
quarterly reflection
I’m quite happy with what I’ve achieved this quarter. My new AI Life Operating System is fun to work with.
Published

July 3, 2025

This is my fourth quarterly review, so I’ve been doing quarterly reviews for a year now. It’s definitely a good thing to think about what I’m doing and do a review. It might even be good to do it every month. I may do that in the future, but not now.

One thing I did this month that makes me very happy—although in some ways it’s quite trivial—was to update my LinkedIn profile. I hadn’t updated it since Dialog Semiconductor was acquired by Renesas, so it still showed me as a consultant for Diasemi. It was a big deal that I did this because I was procrastinating about it. I wrote a little essay on that—Procrastination—along with this image I drew on my train journey:

I used my new LOS “Life Operating System” to help me update the LinkedIn profile. I think creating LOS is perhaps the most important thing I did this quarter. LOS developed from my Tiro recording keeping system idea.

Tiro

I’ve written about my Tiro system a couple of times before:

I keep making changes to it. I now keep the system in Obsidian in five markdown files focused on different areas. This is automatically synced to LOS (via GitHub).

LOS

The idea behind LOS is to have a life and work coach that can help and advise me on where I should be focusing my energies. It is basically a Claude Project with access to a lot of carefully curated information about me. I am still getting together all the information I want to feed into the system. My next step is to somehow extract the most relevant data from the almost 2Tb of work archives I have on iCloud, so that it can see the scope of my previous work.

The reason for the careful curation is because getting the most out of LLMs is all about keeping the context as clean and high-quality as possible. I wrote a bit about that here: Improving LLM responses by cleaning the context

I’ve been really enjoying using LOS. When it came to update my LinkedIn profile, it suggested things I could improve from my previous version. Just the process of creating LOS — getting together past work and summarizing it — has been very useful for me in this respect.

This makes me think — perhaps I should ask it what I should include in this quarterly review. I’ll do that now, it will make a good demo of what I am using it for. (I have used Claude Sonnet for this, as Opus was throwing capacity errors). Here is my prompt:

In the ThisIsJames repository, you will find three quarterly reflections I have done. I am now writing my quarterly reflection for Q1 2025. What do you think I should put in it? How can I improve this process and make it more useful?

In my instructions for LOS I have this line (inspired by a post I read on HN):

Honesty and truthfulness are of primary importance. Avoid American-style positivity, instead aim for German-style bluntness: I absolutely do not want to be told everything I ask is “great”, and that goes double when it’s a dumb idea.

And it seems to have taken being blunt to heart. Which is what I asked for. There is my edited version of its response:

What to include in Q1 2025:

  • Why is there so little business focus? You need to progress beyond your current client base.
  • Don’t just plan steps for business development, actually do them.
  • Do some business networking/outreaching.
  • The above three things are most critical!
  • You set a goal of creating video back in Q4 2024 but don’t seem to have done much. Why not? What’s the real blocker here?
  • Why are you not monetizing your work?
  • You’ve been talking about “innovative visualization” since Q3 2024 but haven’t really done anything.

Process Improvements (ranked by impact):

  • You record metrics for exercise but you need business-related accountability metrics.
  • Set minimum targets: X LinkedIn posts, Y prospect emails, Z networking contacts
  • Your current focus is on learning and technical wins, but light on business realities.
  • Quarterly reviews are too infrequent. Do them monthly.
  • Kill goals that are abandoned (like the Darwin timeline project).
  • Your mortgages and kids university fees aren’t going away — make this visible in your planning.

Bottom line: Your technical skills are solid, but you’re avoiding the hard business development work that your financial situation demands. Stop planning and start executing on client diversification. What specific business development actions did you actually take in Q1 2025?

Ok, well that’s deserved I guess. I did ask it to be a Teutonic business coach. I think I need to take this feedback into account in my plans for the next quarter.

Chatting with it a bit more I realized that I had removed some of the context I had created for it to give me advice. I only found out recently that in Claude Projects, it automatically caches context in GitHub repositories so it’s not necessary to remove it to reduce token count of prompts. So anyway, I added content from the GitHub repository back in and this is how it responded:

Holy shit, you’ve done your homework. Looking at your video ideas analysis - 95+ scored topics, professional LinkedIn titles, clear business relevance - this completely changes my assessment. You’re not winging this, you’ve systematically converted your expertise into a content pipeline.

So perhaps the hard-nosed feedback I got wasn’t deserved after all. I’ve impressed Claude! Yay! [Weak cheers].

Posting to this website

In June I made the commitment to post to this website every day, which I managed to do. Read about it here:

I’ve done quite a lot of posts this month. I’m finding it easier to just write without overthinking things or constantly tweaking stuff. I think one of my favorite posts this month was this one:

Experimenting with Artificial Intelligence

Apart from my LOS system, I did quite a lot of other experimenting and learning with artificial intelligence, including:

That does seem like a lot. I guess the main developments are I now tend to use Claude more than ChatGPT, and I experiment with other models via openrouter.ai using Open WebUI.

Infinite canvas presentation system

I created a new method to create presentations via an “infinite canvas”. I’ve just realized I haven’t written about it. The idea is to combine my notes about a subject with a mechanism to create presentations, so that I can easily turn notes into presentations. I must use the system to create a video. I did write a bit about the idea here (although I abandoned the idea of using a Photoshop file for the infinite canvas):

This is what the end result looked like in that experiment:

I’ve moved on a long way from this and now have a complete working system.

Exercise and weight

I have been using my Tiro data capture system to record my exercise, and my weight from the start of April. Here’s what a typical record looks like:

2025-05-02 08:53:00 rowing:: 8:07 press-ups:: 21, 19, 20 pull-ups:: 14, 14, 14 decline diamond push-ups:: 10, 10, 10 weight:: 90.9

Now I can just copy the exercise log into ChatGPT and ask it to plot graphs. Here’s my weight to date, plus a trend-line to the end of July:

I’ve been trying to lose a bit of weight, just by avoiding carbs, and it seems to be working, albeit slowly.

It’s interesting how bad ChatGPT 4o is at visualizing this data. When I first asked for a trend-line of the weight data, it only used the first five or six points for the line, which as you can imagine gave a ridiculous result. When I asked it to visualize the days I exercised, it created a bar chart with every bar the same length:

This is a rubbish way to visualize the data I think.

Let’s try with Claude. Claude Sonnet 4 approached it in a completely different way, firstly creating an array of the data, then created a calendar visualization. It’s quite good. It’s even given a little analysis of it at the bottom, noting that I had a break due to a cold (which I must have noted in the log).

That’s quite good. Much better than the ChatGPT 4o version. You can see the HTML version here.

Now let’s give Claude a harder task:

Can you now create a visualization of my weight training progression. I change my routine every so often so you'll need to work out how to deal with that in the visualization.

This is what it came up with:

That’s pretty good I think (although it did take a couple more prompts to get to this stage). It’s chosen a good way to represent the different exercises, and added lines when I changed the form of exercises. The key insights are good too.

Overall I’m happy with the way I’ve been exercising and the progress. The big gaps caused by sickness are a shame, but there’s nothing I can do about that.

Meetups

I’ve led sessions in my Discussion Group on the following topics this quarter (I’ve done both sessions twice):

  • Paradise Lost and Satan
  • Masculinity and femininity in 2025

I’ve also hosted sessions on these topics:

  • Technofeudalism
  • Simulating the Universe
  • Hilma af Klint
  • Divide and Conquer - Who Benefits When We Fight?
  • Remedios Varo: Artist of magic and dreams

The Discussion Group is going really well and I get a lot of positive feedback about it. It’s nice to do something that contributes positively to other people’s lives. I definitely intend to keep doing it and improving it.

Creative stuff

I created this strange thing, inspired partly by Paradise Lost:

You can read about it here: Fallen and heavenly angel wheel

Aims for the next quarter

After my blunt review from my LOS business coach I decided I am going to aim to do the following this quarter:

Although LOS retracted its advice after it has seen what I had been up to, I still think these are good practices for me follow.

I’m going to do my first LinkedIn video tomorrow. I think I’ll do it on my LOS system. I also need to use my new “infinite canvas” presentation system to create some videos.