boldra

I’m taking on my first legacy CPAN module to maintain! Jesse has agreed to let me have Locale::Maketext::Simple which I’m very excited about!

https://metacpan.org/pod/Locale::Maketext::Simple

I’ve had enormous benefit from open source throughout my life and I don’t feel like I’ve given as much back as I would have liked. The decision to volunteer for this was partly inspired by Cory Doctorow’s CCC address

I asked ChatGPT to create a bucket list of things I haven’t done yet based on my conversations in the past year, and this was the result.

🧠 Geist & Erkenntnis

  1. Write and finish a small but real book Not a blog series. A coherent essay-book: attention, automation, humans as predictive systems. 120–180 pages. Publish it, even if only via a serious PDF/print-on-demand.

  2. Give a public talk you don’t fully control CCC-style event, university guest lecture, or civic forum. Topic: why “fancy autocomplete” is intellectually lazy. Live Q&A. Risk of embarrassment is the point.

  3. Build a working physical model of an abstract idea Hyperbolic geometry, precession, or attention mechanisms — something people can touch and immediately get. Museum-grade, not school-project.

🌍 Orte (not “countries visited”, but situations)

  1. Spend a month somewhere historically “layered” Examples:

    • Istanbul (Byzantine → Ottoman → modern)
    • Palermo (Greek → Arab → Norman → Italian)
    • Córdoba (Roman → Islamic → Catholic) No rushing. Learn the layers.

  2. Do the Sahara properly, not tourist-style Multi-day, quiet, uncomfortable enough to change your sleep and thinking. You already flirt with this — finish the thought.

  3. Return to one place twice, ten years apart Same neighbourhood, same café if possible. Observe what actually changes: you, not the place.

🛠️ Making Things (you’re a builder, not a consumer)

  1. Finish the mirror maze as a real, operating artefact Not a plan, not sketches. Permits, safety, confused children, annoyed adults. Something that exists without you explaining it.

  2. Create a small, elegant system that runs for years without intervention Software or infrastructure. Boring, robust, documented. The opposite of “startup energy”.

  3. Design a tool that replaces a human bad habit Attention, procrastination, social friction. Not monetised. Just effective.

🎨 Art & Culture (high signal only)

  1. Commission art instead of buying it One serious piece. You argue with the artist. You lose at least one argument.

  2. See one opera and one ballet you expect to hate Then write a brutal but fair post-mortem. If you still hate them, good — now it’s earned.

  3. Curate a tiny exhibition (even informal) Theme matters more than prestige. One room is enough.

🧭 Ethics, Risk, Reality

  1. Put real money into one genuinely long-term ecological rehabilitation project Not ESG fluff. Something slow, legally boring, morally solid. Accept low returns.

  2. Do one thing that visibly scares you but is not dangerous Public vulnerability, not physical risk. If ducks are involved, optional but thematically consistent.

  3. Write your own obituary draft — twice Once now. Once in ten years. Compare.

🪦 The hard one (don’t dodge it)

  1. Deliberately choose what you will not optimise Career status, wealth, reach, productivity — pick at least one and explicitly abandon it. Write it down. Live accordingly.

#6 is the only one where I feel it went way off mark – there are quite a few places I’ve visted over spans of 10 years… I’ve been in Heidelberg for 14 years now. I’ve seen intimate parts of Canberra change for 40+ years. I can even compare corners of London, Berlin, Zurich and Stuttgart over those timescales at a very fine level of detail.

My favourite one is the “opera/ballet that I expect to hate” – I like the framing of it as a challenge and I like the scientific before/after approach.

What was missing most was any reference to my relationships. I suspect this is because I don’t talk to ChatGPT about my friends/family/lovers. That’s private and ChatGPT’s privacy rules aren’t adequate.

There’s not much to tell. I only saw it from the plane.

(not mount Fuji)

Some of my favourite uses of generative AI:

Buying perfume. I'm quickly overwhelmed by comparing perfumes/aftershaves, so I took this approach:

  1. Identified my price range

  2. I took a picture of the shelf

  3. Asked a model to give me 2 or 3 words describing the brand image of each of the products

  4. sampled 3 based on the brand image words, and picked my favourite

Owners association minutes

For looking at residential property to buy, I like to get the protocols of the owners association, put them into Claude, and ask “what should I ask the estate agent?”

Writing an English exam

To compose an English exam, I used this proceedure

  1. scanned in old exams, asked the agent to index them by grammar/vocabulary (giving some examples of the kinds of keywords I wanted in the index).

  2. Wrote a list of what I felt needed to be tested and what I had told the students would be tested

  3. Asked a fresh agent to compose an exam based on the question index and the content I had taught to the required length.

  4. copy/pasted the questions from the input exams to my own draft, with a couple of minor alterations to throw anyone who tried to memorize the exam