Hah, I just realised that even the dashes are inconsistent between emails. 😂

All right, so it’s 11:25am on the day this email goes out, I’m about halfway to London on the train, I have maybe an hour and a half until my Wednesday activities take over.

Last week I wrote that I’d journal what exactly I’ve been doing that I couldn’t get the newsletter written in time. I didn’t, because I didn’t have time to do even that, but looking back at my calendar, here are the rough times:

• weekdays: waking up at 7:30am, mostly work from 9am until about 6pm. This is a large chunk.

• Friday, 3rd, after work: personal life admin: still need to deal with self assessment

• Saturday, 4th: Tough Mudder Midlands 2026. Left in the morning, arrived back home at about 7:30pm, and I was exhausted. And sore

• Sunday, 5th: mostly sore, barely moving, I do not remember much of that day

• Monday, 6th, after work: cutting aluminium, assembling the frame until 2am, then dealing with the cat until 3am

• Tuesday, 7th, after work: went to B&Q to buy cut–to–size plywood to serve as shelves, and go to sleep early because exhausted (cat yesterday)

• Wednesday, 8th: I have the day off, but spending it in London, so getting ready and then catching the train

And here I am.

Here’s what I learned via AI in the last week:

That annoying quiet narration on a lot of tiktok and instagram reels is harmless, but a result of AI

Mostly appears on videos by content farm accounts: they automate the videos and part of the pipeline is the inclusion of that narration. Serves two purposes:

• the audio fingerprinting changes enough that automatic copyright takedowns are mostly skipped, plus it’s now not a repost, but a commentary

• and it’s a byproduct of AI ingest pipelines: they process the videos, describe what it is, put that text as description, and accidentally narrate it too, which helps with the algorithm picking it up

It does not hijack your phone’s microphone to serve you ads of different topics. The video itself and the platform would be so much easier for this, the mic is not turned on either way.

Phones usually run AEC – Acoustic Echo Cancellation

Phone knows what the sounds are that it’s playing through its speakers.

Therefore it can process out that sound from the incoming sounds by the microphone.

Which also means any narration on a video that’s audible wouldn’t get past the AEC in the phone.

Different body parts have different radiation tolerance

Your eyes can take 20mSv/year, whereas skin / extremities can deal with 500mSv/year (milliSievert per year).

For context, the background radiation is around 50nSv/hr (nanoSievert per hour), which would come to roughly 0.5mSv/year. The limits are always “in addition to the background radiation".”

Fun reading in case you want to dig in: https://www.legislation.gov.uk/uksi/2017/1075/contents.

(I have not yet, it’s 11:43am, ~3 hours before the email goes out, and I’m not done yet!)

On MacOS, zsh has so many more files than .zshrc

The load order is .zshenv → .zprofile → .zshrc → .zlogin.

I started to separate out what I put in my .zshrc into these different files, starting with all the export VARNAME=”value” lines going into .zshenv.

TIG welding means Tungsten Inert Gas welding

Tungsten is non-consumable, it just gets really hot and melts the material. If you need to add material, that’s a separate rod from your other hand.

Inert gas, Argon usually, flows around the weld area to keep oxygen out.

It’s different from MIG welding.

MIG welding is Metal Inert Gas welding

This one has a continuous spool of consumable metal that becomes the filler material. The gas is either a CO₂/Argon mix, or just straight CO₂.

That’s it for this week! I’m in time, going to tidy this up and schedule it. 🙂