Thoughts and reflections from the past week or so from my own financial independence campaign.

Progress on my goals

Boat life

This week I services the Jabsco marine toilet. Smelly but necessary. Lady SierraWhiskyMike made sure to keep right out of the way for that one!

The servicing kit cost £35, which is OK I guess. I can also confirm that despite what Jabsco says, the modern servicing kit does fit on the older toilet mechanisms, but obviously you can’t take advantage of a locking flush lever catch it your sea toilet doesn’t have one – like ours.

So I now I know a lot about sea toilets. Great. I’m sure it will make for riveting conversation at polite parties.

Start-up work

Hoping to close our investment next week. Terms are basically agreed now, so this weekend I need to amend the articles of the company and draft some board resolutions to approve the investments.

Tedious yet important work!

Distractions and detours

Bitcoin price dip

The price of a full bitcoin dropped 30-odd percent this month so I put a bit of my cash into it. I figure that it’s unlikely to drop another 30% any time soon, and it’s a lot faster to cash out of than equities so if I need the cash to live on I can easily sell it or spend it.

I’m still comfortably up on my position from a couple of years ago and I get paid in BTC, so my setup for spending it on a Visa card or cashing out to sterling is fairly streamlined. I get that most people don’t have that kind of luxury because bills are in fiat and for most people a mortgage/rent is their biggest expense, so operating through Bitcoin isn’t a viable option yet.

Even if this fails and the price of a bitcoin drops another 50% I’m pretty comfortable.

“AI” experiments

I’ve been dabbling with “AI” products for some of my first draft work. So far I’ve tried:

  • Elephas, a MacOS app that interacts with several LLMs and works offline with data “brains” of information you curate
  • LLM Studio, which is an operating platform where you download LLMs to work locally – I’ve used Qwen and LLama models
  • ChatGPT Plus

I’m currently paying for ChatGPT Plus and Elephas.

Experience so far

LM Studio is… OK. I used it to help me summarise CVs for our business plan and Llama did the first draft pretty damned well, speeding me up. The models I run locally are great for things like copy or summarising documents without sharing information around to LLM providers or searching within documents, but Elephas can also do that and it’s a little bit smoother to use. Then again, Elephas is £20 a month whereas LM Studio is free…

Elephas is a knowledge retainer. It does OK at drafting, but the real use is that you can deliberately limit the “brains” of documents it uses for context, and then when you ask it a question it picks from an aggregate of LLMs to answer you. I started off with Elephas and I can see how a law firm could use it for knowledge retention and case law analysis, but like all LLMs it’s not super reliable and I’ve had to direct it to documents in the “brain” that it doesn’t seem to want to read.

ChatGPT is very hit-and-miss. I like using it as a research agent and it’s exceptional at running logic based on evidence, but it’s not the most accurate thing ever and you have to constantly challenge its assumptions. I’d say it’s worth the money if you’re using it to help build your own thoughts and bounce ideas around, but never trust it for decisions or research into things you don’t already have a good background knowledge of. It’s also really good at helping you with computer software issues: for example, it’s teaching me to do things on MS Word that my last employer bought plug-in software to do but is actually easily set up as a style or template.

Thoughts to far

I don’t think the robot uprising is any time soon. None of the “AI” tools have been worker-replacing so far. However, I can see that my productivity has improved by integrating some AI systems into my MacBook in carefully selected ways.

I’m not sure moving forward if I shall keep LM Studio and the LLMs I’ve downloaded then get rid of Elephas, or if I keep Elephas and abandon LM Studio since I basically do the same things on them. My gut says that ChatGPT is working for me for research and that if I can sort out how I use LM Studio then I can get rid of Elephas entirely.

Further rabbit holes

I think I’m going to look into creating AI agents for work when the first set-up rush of getting the company trading is out of the way. I can see the value in automating a lot of routine tasks and for analysis of data inputs. Getting Copilot for our business is going to be an affordable expense, so I’ve got something to play with that could cost us very little.

What I’d like to do is turn myself into an AI-augmented lawyer, so that I can maximise my output across my projects and free up thinking space (and time!) for my hobbies and my home life. It would be dreamy if I could create precedents that I’m happy with and get AI agents to pull together first drafts of stuff for me to review and edit. This would let my business ideas scale without a fuller legal team or too much outsourcing to law firms of stuff that I should be able to do in-house but just lack capacity for.

Jeans rabbit hole

One of the things I’ve been thinking about is how quickly I go through jeans, which is important now that I basically live in the same 2-3 pairs of jeans for work and home.

I usually buy £26 M&S stretch jeans and they typically last 6-9 months. I’ve had pairs make it a year but that’s usually if they’re black: I tend to wear shades of blue more often.

Anyway, on a “Buy It For Life” group I saw a post about raw denim that led me down a rabbit hole on how jeans are made and why my jeans don’t last as long. Short version: stretchy jeans are lightweight, but you can still buy traditional weight jeans that are usually 50-100% tougher material.

The downsides? Well, cost. Also, you have to break them in, because they’re pure cotton and don’t contain any elastane. Oh, and you aren’t mean to wash them for a couple of months while you break them in.

OK.

So I I had a look around at “raw denim” and “selvedge denim”, which is apparently better material because it’s made using the traditional old style loom. I… don’t really care about that. All I want is trousers that I can live in that don’t fall apart every year. Raw denim and/or selvedge denim (there’s some nuance here that I didn’t understand) sets you back £120-500 a pair, but boasts that they should last 3-5 years easily with daily use.

That’s cool, but also expensive. If I was still in law? Screw it, I’d give it a go. This year though, while I’m underpaid? That’s a bigger ask.

I used ChatGPT to have a look at this. While the cost:benefit of selvedge jeans might make sense because I live in two pairs of trousers and they get worn on a boat, the maths suggests that actually I might be better off in workwear jeans because those have the same quality material despite being made on modern weaving looms. Like I said, I really don’t care at this point about fashion per se, I just want reliable hard wearing kit that I can use for social situations and a casual business meeting.

Anyway, turns out workwear jeans that look like normal jeans are about £80, which is much more affordable.

I’m going to try a pair and see how I get on.

Non-FIRE goals

My guitar rock god quest (AKA learning to play)

My guitar teacher’s partner had to go to hospital for an operation this week, so we cancelled guitar while he brought the kids to go see her. Rightly so!

Orion is sort of coming together.

Start-up life is seriously impacting my practice time, which I’m a little conscious of. When the investment docs are signed, some of the burden falls away from me for a while and I can manage my time better.

Well, that’s the hope!

I’m still doing guitar and enjoying it though, which is the important thing. I’m getting better even if I’m not getting consistently good yet.

Fitness

I completely neglected to get the physio sorted like I said I would, so I’m now booked in for tomorrow afternoon to get a consultation having booked it in a panic on Saturday morning.

My knee isn’t really getting better beyond an “it doesn’t hurt right this second” point. I had to rest this week to see if it would calm down and I didn’t think CrossFit was a wise decision if I want to continue walking… which I very much do!

I’m a bit nervous about what the physio might tell me. It’s like I know conceptually I’m nearly 40 and spent my twenties getting thrashed in boots and weighted kit, so obviously there will be a price to pay for that; but emotionally I feel like that was only a short time ago, even though I’ve been a civilian for over 6 years now.

I’ve also found that I can get a heavily discounted gym membership at the leisure centre including a swimming pool membership, provided that I don’t go between 1700-2000hrs midweek. Being effectively self-employed, that seems surprisingly doable. I might go with that, so if my knees are shot at least I’ll have cardio options that aren’t running.

Final thoughts

It’s been a week of investigation and diving into things.

Obviously I’m keeping the work stuff light on here because it would be too easy to connect this blog to what I do. I’m not secret, but I like to be a little discreet.

The AI / LLM thing is fun. I’ve definitely seen value in LLMs for churning through reams of bureaucracy quickly, so I think I’m sold on potential utility; but I need to refine how I want to use this technology in practice. It won’t replace people, but it could definitely give me increased effectiveness at scale.

The ongoing knee injuries suck. I’m hoping the physio will do a scan or something so I can plan how I’m going to keep active for the next 40 years or so. Can’t do nothing forever, but also don’t want to keep trying high impact stuff if that’s likely to leave me as a cripple in my old age. For both of these, the reason is the same: my healthspan is important to me.

I’ll keep you updated on the weird jeans detour.

My financial independence campaign continues!