This week is heavy, I have about 30 responsibilities, GitHub Copilot revamped their billing model, I have absolutely no idea how much my work is costing the company, and there’s the conflicting issues of needing to do work, and going to a conference at the same time which results in sleeping maybe 2 hours.

This issue does contain a short-ish article that’s different than the usual tiny points of stuff. I’m curious what you think about it. Please let me know!

Others I learned this week were mainly around starting a watch microbrand, figuring out the purpose of a chain I got with the grandfather clock I bought, garage door operators vs smart home devices, and tax implications of buying an EV via a business vs personal.

_artikel: I am convinced people are holding AI wrong

I’m not going to edit this much, but basically I keep reading and watching all the news on how developers chew through thousands of dollars of AI token usage a month! Or a week. Or a day! This absolutely baffles me, as someone who practically lives inside several AI harnesses every day. On my personal Claude Max5 subscription I can’t seem to reach the limit of AI usage, even when I use Opus 4.8.

To the folks who run out of money each month: what are you doing? What are you getting the AI to do for you? It seems incredibly wasteful. And again, I don’t know whether it’s me being an absolute n00b at handling AI, and I just haven’t levelled up to power user levels, or I’m actually a pro at using AI, and I get a lot of stuff done with comparatively little amounts of token usage.

Holding AI correctly is like holding a knife correctly: if the knife is sharp, and you know what you’re doing, you can dice a lot of onions in a short amount of time. If you’re using a spoon and you’re full-fist grabbing it, dicing onions is going to be really difficult. Are the people running out of tokens trying to use spoons to dice onions? To translate, are they using AI for things AI wasn’t meant to be used?

“Make the UI look good” is not a great prompt.

I am trying my hardest to not be judgemental about it, but I also see how all of this will come off as judgemental. If anyone’s willing to show me around their AI usage where they run out of money or rack up a huge monthly bill, invite me! I’ll sign all the NDAs you want, but I have to know what’s happening there!

Okay, back to the things I learned:

Starting a watch microbrand is surprisingly straightforward

Most microbrands apparently use the NH35 / NH36 Seiko Instruments movements to power their watches. It’s off–the–shelf, serviceable, easy to maintain, not super expensive, very reliable.

Then you have cases you can buy in bulk fitting those movements already. Same with crystals, straps, and hands.

There’s an entire ecosystem built around those two movements.

Watch dial printing is an art form that’s also commoditised

Similarly there are a few companies that offer either off–the–shelf dials for the NH35/36 movements, or they allow you to upload your designs and print them for you. See https://www.dialmaker.shop/collections/custom-dial and https://deklawatches.com/en/dials and https://www.soflypart.com/product-category/watch-parts/dial/.

The microbrands are hand assembled apparently

Once you order everything, all you gotta do is put everything together. It’s still going to be a lot of work, but not nearly as daunting as having to buy a six axis CNC mill / lathe and learn how to CAD / CAM.

That also means that being unique, or having a good reason for customers to buy your watches is that much harder.

Don’t try to make watch hands

They are tiny, you need specialised equipment, skills, and minimum volume to make it worth it. Buy them.

Desktop CNC milling machines can’t do steel or titanium

Especially not the 316L version of steel. The reason comes down to:

• 316L steel is very hard

• tool needs to spin fast to remove material

• tool also needs to be push on steel to bite and remove material

• machine therefore needs to be heavy

• and powerful

• desktop machines are neither heavy nor powerful

• and thus they don’t spin fast

The end result is that a Makera Carvera Air would struggle with anything harder than brass or aluminium.

Most garage door openers have pins that can be used for remote control

For example a Shelly Plus 1 (£12.49 at the time of writing after the £4 discount) can be wired into the pins of practically any and all modern garage door openers and bam, you have a garage door you can open with your phone without having to buy a system that’s proprietary and expensive.

Home Assistant is your friend here.

SOMMER and Marantec garage door openers are perfectly capable even at the entriest of levels.

Buying an EV via a company and allowing personal use is probably more money-efficient than buying an EV personally

This is not tax advice, go talk to your local qualified and regulated tax person before deciding on anything and have them run numbers for your circumstances!

Assumptions: worst case scenario, person is in 45% tax bracket, no tax allowances left, £40,000 car (+£8,000 vat), 4 years, £20,000 resale value.

1. If you buy personally, you need to use money you extracted from the company and paid dividend tax / paye tax on. That’s after the company paid corporation tax. And you’re personally responsible for running it. Total cost comes to £93,867 (ninety-three thousand, not a typo). If you sell the car, that goes down to £73,867, no tax on sales of car.

2. If company buys, £46,580. If company sells car, it needs to pay 25% corporation tax on value, net proceeds becomes £15,000, so total cost is £31,580.

3. If company leases, assuming £6,000/year ex vat, £28,980 total for 4 years. No sales because never owned the car.

The difference between personally owning vehicle and company owning a vehicle is wild. Don’t want to put detailed calculations, because this is supposed to be a list of quick learnings email and it’s already super long!

The chain is useless

The Kieninger clock is entirely cable driven, has three weights, and the chain is a short 60cm link of rings with no open rings or hooks on either ends. There’s no purpose for the chain as far as I can tell. I’ll ask a horologist though Soon™

This has been issue 06. I’m going to sleep, it’s 2am, and I need to be up in 3 hours. Fun things all around!

As always, if you liked this, tell me, email me, send me a message, share with friends, all the usual bits. 🙂

Loveyoubai! ❤️