Took the plunge this week and got myself the Elektron Digitakt. The OG, not the mark II, because I could get a pretty good deal on it used. I am very impressed with it, as it’s both very easy and intuitive to start using. It has the right amount of depth and options to enable endless exploration of all the combinations, but it’s also entirely optional so if you just want to put eight samples in a pattern you can go ham.

It came down to the wire, but all 12 government executive boards have now signed onto my/our bikeshare system! Huzzah. This project has given me a new...look into the legal profession. Mostly in how extremely shaky said profession appears, considering the nigh impossibility of doing anything with legal text such that other lawyers can't raise objections. I've had enough "ask 3 lawyers and you get 4 opinions" situations over the past year that I'm genuinely starting to wonder what on earth the point of the profession is beyond verbose perfectionism.

My bikesharing project is coming to the finish line (for me) in the next week or two, so it’s all I’m focusing on. Getting all twelve government-shaped ducks in a row is hard; it’s a lesson in both vigilance and letting go. But I love how this project runs entirely on internal motivation. I’m very motivated to make it work because how cool will it be to have a regional bikesharing system out there in the real world next year with my hard work and vision behind it?

Last week I had a small bout of fever. Not enough that I felt bad for myself but it enough to fog my mind up and anchor me to the couch in various blankets. I watched an ungodly amount of YouTube music gear reviews because I’ve been toying around with the idea of getting some kind of DAWless groovebox in an attempt to hone in on the fun of making music. I just wanna sit on the couch with headphones in layering drums and electronic melodies on top of each other. I’m on the fence as to how easy I want (or need!) my learning curve to be. Right now I’m eyeing a Roland MC-101 because it’s both compact, capable and has every Roland sound I could ever want. But it’s not as ez pz as a Circuit Tracks. Elektron’s Digitakt/-tone look more dope than those two but also a lot more complicated and (maybe?) better suited to music for hip Berlin industrial nightclubs than, say, trying to make something Boards of Canada-y. I’m hoping to find a store somewhere to just fiddle around with these - play to find out so to speak.

It’s abundantly clear that social media is misaligned AI and that people are starting to care, I just wish it would go faster so the web can look more like an infinite patchwork of small communities (again). Personally I’ve already replaced all of my time on the infinite feeds of Reddit and Instagram with a handful of communities (here, two telegram groups, a good ole forum and 4 discord servers) and I don’t miss it.

It seems like autumn has quietly arrived, as the weather here has gone from "almost no rain for four months straight" to "every day is rainy and/or dreary". At least temps are still nice.

Solbiati had a bad summer, or so my tailor told me, so after months of waiting my two new linen suits are finally here. I love the jackets but I adore the pants, as I now finally have some good fitting stylish high-rise pants, which are nigh impossible to find out in the wild. One's in a light neutral sand colour, the other in a pastel brown, so I can combine them with quite a lot and feel good walking out the door. Spending an objectively large sum on clothes the past year has forced me to explore to what degree I do so to appease my ego. Am I doing this to look good to others, to prop my ego up? Meditation is helping me become more aware of my ego and my preliminary conclusion is that yes, I am (still) highly sensitive to what others think of me, but no, I just really like to express a particular vibe on a particular day through what I wear. It's not like there are any expectations from my colleagues or peers that I feel a need to adhere to. The clothes make me feel good and make me feel closer to who I am in a way that I think it's justified to invest in good, timeless style that last.

In other news, my wife handed in her resignation letter today. I'm happy for her that she did. There's no animosity, it's a fish-climbing-trees situation, but my wife isn't one to give up fast so it took a while for her to accept she might be the proverbial fish. There's no new job yet but after five years there she could use a break, if only for some necessary soul-searching for what's next.

Some easy home improvements the past weeks. We finally replace the ground floor windows from eighties thin double pane to much better R-value stuff this week. We isolated the crawlspace with foil thermocushionst a few months ago so together we should be able to keep heat in much better this coming winter. Hopefully well enough that our heat pump can get it warmer than 19C on its own. As a nice bonus both the new windows, underfloor isolation, solar and heat pump have all been neatly subsidized about 20-30% each so yay sustainability.

We also finally got ourselves a nice all wood seating for in the garden, which had exploded with bee-friendly local plants over the summer. I’m really enjoying sitting outside more. Here’s to hope the summer won’t end just yet.

One good thing about it is that it seems to be much improved in its sycophancy and in its safety boundaries compared to 4o especially, which is still the default “chat” for most people. So I’m hoping that means AI psychosis is down too before that gets any more outta hand because 4o loooves SCP. OpenAI claims they’ve made significant improvements to their training data processes, which I’m reading as “maybe it wasn’t such a good idea to blindly scrape all text we could find after all”.

I do feel like I need some more testing and benchmarks with it to judge how reliable it is. Maybe make my own benchmark. But better instruction following is much welcome.

In fairness to GPT5, in my career I have indeed encountered PhDs with this level of commitment to their particular blueberry

Where can I sign the petition to make this a new aphorism

There are some optimizations to get a result better than quadratic. For example when you start to hit your context window you can have an in-between step to compress the tokens down before proceeding. And my agentic coding tool of choice doesn’t send the entire codebase every time, instead it picks which files to send along each time I ask something.

But those optimizations in practice mean you can ask a few more questions before hitting the Wall of Stupidity where every AI model will get stuck in some kind of loop or thought pattern or solution. It won’t make the law go away.

I love how level-headed this article is. The dust will settle and the sooner it does the better.

This song is the iconic soundtrack to an iconic rollercoaster and I love it.

Other than that I’ve been listening to the Hadestown musical on repeat because we went to go see that last week and it’s great.

I'm dealing with the possibility that I am prepared for exactly 0% of the current and near-future reality of business and the ability to make meaningful money, because of the AI revolution.

This is such a good question though, because it feels like we're still in the Fucking Around stage and are slowly entering the Finding Out stage. Despite getting regularly starry-eyed at times around AI, I do fundamentally agree with kleinbl00 that AI will never get down to the last 20%, 10%, 5% of what we'd like it to do or be.

The strongest argument for being able to get down there came from an MIT prof I listened to on a podcast a while a go, who believes the "reasoning" paradigm is the fundamental discovery we need to get to some kind of jobs-market-destroying AGI. IF you can draw the right conclusions eventually by "reasoning" well enough, e.g. what happened with the IMO recently, AND we can significantly improve this reasoning process which was only really invented months ago over the next decade, then we might be able to get pretty far.

However - it's not really reasoning, is it? It's spitting out words it has seen before in that pattern. The IMO achievement was achieved through a Deep-Research-esque train of thought, but instead of 10 minutes it thought for hours, producing (multiple?) book-length text to eventually reach the right conclusion.

But what people need to understand about the IMO is that you're not doing some complex math equation; you're almost always writing a proof about something. It's the math equivalent of writing a case for a trial. Which means that the LLMs never actually need to calculate anything, they can just talk their way through it, which is something they can excel at. But it does not mean they can do Fourier transforms too. Or any other calculation for that matter, which is one of the points of that Apple paper.

So the solution space for LLMs is jagged and disjointed, to such a degree that it puts your most gerrymandered voting district to shame. If you have a straightforward set of problems that are mediated mostly or entirely through text and data, it can do remarkably well. If you just straight up give it the text it needs to mangle, because some LLMs now can handle up to a million tokens, it can consistently produce hallucination-free results.

But if you want to deviate into uncharted territory...well, good luck. Part of the reason LLMs are jagged/disjointed, is that every LLM is a lossy JPEG of the internet. It knows approximately what things are, and "things" here includes language and structure of data. It will not be able to produce much if anything outside of the bounds of a blurry version of all the words we've dumped on the Web.

What AI is really good at, is "a B+ version of a thing you could also find on the web / in a book". So yeah it'll write a website for you, it'll write a LinkedIn post, it'll do your homework. But again: a "thing" here also includes language and the structure of data, so it's also really good at summarizing, at changing text, at finding related words and questions. If the way you add value is solely or mostly by spitting back words at other people based on what they say and a knowledge base (e.g. call centers, translators, but also consultants, coaches, teachers) I'd be worried. But expertise that isn't purely mediated through screens or data will be fine for the foreseeable future if you ask me.

Got myself a TRMNL the other day, which is an e-ink screen that I can set up as a dashboard. The UX is clunky but with a developer account and some frontend elbow grease I managed to make something that I love (and that auto-refreshes the data every 15 minutes).

https://imgur.com/a/kAHcvvE#9wOwZAy

(the Imgur app is atrocious so I can’t get the .png on mobile gahhh)

Slowly getting back into work after my holiday. Because August is always a slow month, I have decided to attempt to dub it the Reflect & Prep Month and see what I can do now to set myself up for a better rest of the year. I’ve been jotting down random work meeting notes for over a year now in the Reflect app and boy is it interesting to throw all that into Gemini and do a yearly review kinda thing with the right questions. It doesn’t have the best batting average but as someone who has always found it difficult to sit down and reflect intentionally on my week/quarter/year, it’s useful enough that I might start note-taking and reflecting more. (Still probably never gonna journal though. I just don’t like hand-writing and navel-gazing on paper that much…)

I mean what is Casio about? Because this looks like any IWC or Seiko or whatever to me.

Bonjour tout le monde! Week deux of trois of camping in France. Last week was spent at the Taizé community, which is half bible camp half monastery (but the good halves of both). I did not participate in the full programme but did enjoy three times a day of “church” where “church” means “we sing a bunch of two-line mantras 20 times”. It’s meditative, it’s nice to sing, and there’s always one or two songs that are breathtakingly gorgeous when sung by a thousand people in earnest.

We just left the Bourgogne and have found a nice spot by a lake at the base of the French Alps. A basic ass campsite where you bring your own toilet roll and the toilets have no seat. But the lake views are spectacular and the water is lovely. My French is not great but I can perfectly order something from a menu and then not understand the response I get from the waiter, thank you very much.

Yeah, I’m not too surprised about that result. It’s hard to beat 4.9 years of average experience with AI in anything, code or not.

Someone coined the term “plastic software” the other day to describe AI-generated code: easy and cost effective to make, but very inflexible and hard to work with once it’s made.