I don’t mean that in the sense that I’m addicted to it. I literally can’t turn it off. When I shut it down, it immediately boots up again. When I tell it to hibernate, it immediately boots up again.

From Microsoft:

Symptoms

After installing this update, some Secure Launch-capable PCs with Virtual Secure Mode (VSM) enabled are unable to shut down or enter hibernation. Instead, the device restarts.¹

Next step

This issue will be addressed in a future Windows update. Additional information will be shared as soon as it becomes available.

How has it come to this? Technology used to be a source of optimism. We used to dream about all the ways it could improve our lives. Now we’re lucky if a system upgrade doesn’t break the fucking power button.

You know how it’s pretty easy to spot an online post or comment that was written with ChatGPT? It’s not that they use em-dashes—plenty of great writers use em-dashes—it’s that they have a general air of disconnectedness and they write with a patronizing, prescriptive structure. It’s easy to spot most AI slop.

However, I am cognizant of survivorship bias. There is a lot of AI-generated content that slips by unnoticed. If you have a strong command of both your subject matter and written prose you can prompt your way to a high-quality output, but in order to do that you have to give it more than it gives back. It’s a lopsided, toxic relationship, and these days there are a lot of lovesick people out there.

Unfortunately, your boss is one of those sad saps standing on the curb with nothing but a bouquet of flowers and a dream.

In my mind there are three different types of tasks that we might get a Generative AI tool or “Agent”¹ to do for us:

1. There is little disagreement that Generative AI is bad at anything that might loosely be described as art, which I’ll define for myself as anything unique that is created with a mixture of curiosity, imagination and craftsmanship.
2. Generative AI is great at tasks and processes that require no imagination, like extracting structured data from written text, summarizing a transcript, or repetitive clerical tasks.
3. Generative AI might be good at domains like programming or law, which can span a massive spectrum between the utterly routine and the very incarnation of elegance.

The third category is the most interesting to me. There are wildly divergent opinions on the utility (or futility) of Generative AI for programming. The world of bits and bytes can be intractable because what you build is limited only by the guardrails of your imagination. The substrate is what you decide it is. The raw materials are undefined until you declare them.

In the physical world of atoms, we have a general idea of the forces, chemicals and components that we need to build a bench, a bridge or an airplane. That isn’t to discount the expertise required to build those things, I note only that there is an inherent tactile structure to building things in physical space that doesn’t exist in the digital one. Things fit together, or they don’t. Most failure modes are known in advance, unlike the sly bugs that lie in wait within every digital codebase.

Because it is not grounded in the physical world, software complexity fits a fat-tailed distribution. A lot of applications follow the tried and true recipe of reading and writing information to a database with a bit of business logic and reporting mixed in. The techniques for creating such an application are well documented and well understood, and this is the sort of thing that Generative AI can do very well with little correction or oversight on the part of the operator. For this reason, prototyping greenfield projects is a great use of the technology.

As we navigate further to the right on our fat-tailed distribution of software complexity, we quickly enter a different world. Any piece of software is comprised of layers—each layer building upon the layer below it, disconnected from and naive to the specific implementation details underpinning it. This is called abstraction and it’s a foundational concept in programming. In simple software we might be coding at just one layer of abstraction above a well-documented framework that the Generative AI is pre-trained on. But in complex and/or mature software, we’ve already written a few layers of our own, and the AI hasn’t been trained on the building blocks we’re asking it to use.

I believe most of the disparity in lived experience with Generative AI among software developers can be explained by plotting their work on this approximate log-normal (fat-tailed) distribution. I suspect this shape holds true in many other professional domains as well.

It is possible to get a modern Generative AI to work on complex projects in the tail part of the graph, but it needs much more attention and guidance. That is, the operator needs to know both the subject area extremely well, and the nuances of the Generative AI tool extremely well. Obtaining a positive outcome at this end of the chart is a difficult skill to master in itself, with novel risks and pitfalls. In many instances it is faster and less error prone to write the code yourself, since you will need to step in and course correct the AI a lot. There is ample value in certainty, and you forego that at the outset. If you don’t have the expertise to identify its missteps (and there will be many), then you are firmly out over your skis. You can see how frustrating this is for everyone involved in real time: take a look at Microsoft’s own engineers losing their minds trying to get AI to write correct code for them (also here, here and here).

Your boss has been sold a grand lie. It’s not their fault. They have been set adrift in a sea of misdirection, misaligned incentives, grift, absolutism, desperation, and stupidity. I have never seen a manic hype cycle like this one and neither have they.

ChatGPT came along just as 15 years of free money dried up, leaving an overweight tech industry clamoring for something, anything to keep the capital spigot flowing. Not only could Generative AI create an entirely new class of products and start-ups, it could also be used as cover to lower headcounts and put those pesky, overpaid software developers back in their place.

So the salespeople started buying seafood towers and the scrum masters scrummed with renewed vigor, and every downstream middle-manager through C-suite executive were convinced that they didn’t just want AI… they needed AI. There’s an AI button on your keyboard now. notepad.exe has Copilot built in. There’s no budget for anything without AI. If you don’t use AI, don’t bother showing up on Monday.

To state the quiet part out loud, the promise of Generative AI is “efficiency”, and “efficiency” simply means doing the same amount of work with fewer people. So if you came to this post wondering if your job is at risk, the answer is probably yes, but not because of AI—because your boss has been pumped full of hot air.

Your boss has been told that an AI “Agent” is equivalent to a person. That you can set it on a task, and with enough iterations of the loop, it will arrive at the correct solution/feature/output. They’ve been told that their competitors have many such tools and “Agents”. They’ve been told they’re falling behind. They can’t know it for sure, and they can’t dispel it, but everyone is saying it, including their own boss, and their boss’s boss, and those well-dressed chaps on that panel last week. It’s not your boss’s fault, they just have to to keep up with <dastardly competitor>, who no doubt is using Workday: the powerful AI platform that keeps your most important assets on track, every decision on point, and your fleet of AI agents at peak performance. That’s Workday.©

Jesus Christ.

The reality is this: in order to apply Generative AI to a task, there needs to be a human operator in the loop who understands that task BETTER than the AI. The classic adage of “garbage in; garbage out” applies. You cannot take someone who lives on the median of the distribution in the chart above, give them a replit account, and expect top-tier output.

The upper bound on what your company can accomplish with Generative AI is the level of your most proficient colleague. It is as true in 2025 as it was in 2020 as it was in 1820: you cannot do great things without great people. What you can now accomplish are more middling things with the same amount of great people. That has business value, to be sure. There is plenty of middling work that needs to be done. To quote my man Alec, “world needs plenty of bartenders!”²

Now, I have no empirical basis for these numbers aside from my own experience and intuition, but if I’m being VERY generous I would peg my own efficiency improvements using Generative AI at somewhere around 20% to 30% on average. On prototyping and side projects, I’d guess that I’m approaching 100%, but that’s not real work. On my mature main project and its components, the improvement is well below 10%. Your mileage may vary depending on your mix, but it’s not a human ass in a human seat, that much is true. I would estimate that an AI-forward company could drop a single junior developer for every 3 or 4 senior developers, which would be on the order of a 10% to 15% reduction in compensation expense in the base case of a company with only 4 or 5 employees, all but one of them senior. The savings would be well below 10% in a larger, more balanced pool.

How might you protect yourself from the whims of your stupid, gullible boss who hasn’t been enlightened by my napkin math? This hype cycle will crash in due time as they all do, but Generative AI isn’t going to uninvent itself. Our environment has indeed changed, so it’s time to adapt.

If my assertions prove correct, it will be the mundane work that ultimately gets carved out and handed off to an AI under the supervision of an overqualified operator. This runs counter to the current narrative that you can have average people accomplish above-average things with AI, but I don’t think that’s how it will shake out (much to the chagrin of MBA consultants everywhere, themselves rather average).

If you’re already in that “skilled operator” category and using Generative AI to expedite your more menial work, continue advancing your ability to use the tools but remember not to let your basic skills atrophy. I can say my spelling is far worse now with the ubiquity of spell-check than it was in grade school. Likewise, how many people can still parallel park under pressure without a backup camera, or navigate across town without GPS navigation? Take your cue from the aviation industry where pilots regularly fly manually despite auto-pilot to keep their skills up.

If you’re starting out, that’s a trickier spot to be in, but it’s not impossible. Lean into fundamentals and keep your pencil sharp, because there will always be a place for the person who actually understands the code that is deployed, and that will only amplify in the future. Your colleagues or classmates are all learning to ride a motorcycle before they know how to pedal a bicycle. Learn how to pedal the bike.

I stated early on that art is safe. I believe that will always be true because art is how we express humanity. It is the antithesis of machine. There may be no artistry apparent in your day job, but curiosity, craftsmanship and imagination are the ingredients of mastery no matter what you do. An AI agent cannot replicate your taste. It has none of your flourish or flair. It has no style. It is not cool.

If you’re a programmer, recognize that programming is design. If you’re a labourer, approach your job like an engineer. If you’re an engineer, approach your job like an architect. Carve your name into your work and you’ll be alright.

Part of the Far West 2024 series.

I’m a member on an old-school overlanding forum called Horizons Unlimited which is full of folks who’ve ridden their motorcycles to every corner of the earth. When I threw my itinerary for Far West out there to seek advice, one of the old-timers opened his comment with this:

“I’m having some difficulty imagining your trip as it’ll unfold on the ground. You’re covering a lot of ground, but don’t appear to be actually seeing very much, mainly because so much of what’s worth seeing along your route requires either long side trips, getting off the bike and walking around, or both.”

— markharf, The HUBB

I responded to the effect that I knew it was an aggressive itinerary, but three weeks was all I could pull off and that I’d have to make do. It was a motorcycle trip after all—shouldn’t that be the main activity?

Ummm, no. Not really.

Lesson 1: Slack

To qualify this a bit:

1. I still had a blast, and
2. the somewhat impromptu 3 day hiatus in Palm Springs compressed the riding schedule to its limit, and I chose this tradeoff knowing it would make the ride more challenging.

But the wise man was right. I spent a lot of time riding past places that I would have loved to have spent a day or two exploring. The motorcycling, fun as it is, is not actually the point. Moreover, I didn’t leave any slack to accommodate unexpected issues like not realizing my chain was about to snap until I was stuck at the Grand Canyon. Oh, mistakes were made.

If I subtract the three days in Palm Springs, I covered an average of 380 kilometres per day. That doesn’t sound like much in the context of four-wheeled road trips, but they are big days on a bike when you’re avoiding the freeway. My Garmin estimated 5-7 hours in the saddle most days, and that’s before you factor in a fuel stop or two, lunch, points of interest, trying and failing to find a god-damn banana over and over again, and coffee breaks. There is also hard work involved—packing and loading the bike, setting up and tearing down camp where applicable, bike maintenance, sitting around in laundromats, on and on. It was easily 10 hours of logistics each day. Sustainable for three weeks with a good break in the middle, but I wouldn’t do it this way again.

On the next trip, I will cut the target mileage-to-time ratio in half. 200 km per day average on an asphalt trip, less if aiming for dirt, which is the other thing I would have done more of if I had had a more relaxed time budget. If you’re using this as a guide it’s important to note that these are overall averages, but I would aim to ride only 4 or 5 days each week, so the actual mileage on a riding day would be higher.

Lesson 2: Gear, parts & packing

I did alright in this department. You can see my packout in my earlier post: Packing Lucia for a four-season motorcycle camping trip.

This all worked out okay. I can’t say enough about how solid Kriega bags are, and the best idea I had was to stick a mountain bike handlebar bag on the bars as a quick stash for random stuff like ferry tickets, GoPro batteries, cash/change, and ear plugs.

Next time I will look to swap out the 12 litre side saddles with the 18 litre variants that mount to the same base, as that is free real estate as far as the ride goes. I will scrap either the 10 litre or 20 litre tail/tank bag in favour of a 9 litre Kriega backpack. There is value in keeping your most important possessions on your person, and a camera mount on the shoulder strap would have been perfect.

The shoe situation was pretty dumb. I ended up bringing mid-top Arc’Teryx hiking boots, but I will invest in a more versatile and less bulky pair of all-purpose trekking shoes on the next trip. I felt quite goofy wearing hiking boots while eating a $55 scallop risotto in Palm Springs when we went out for a nice dinner.

Other minor adjustments I would make:

• More compact and better quality cookware;
• Smaller fuel canister;
• Small size, leak proof containers for oil, spices, etc;
• Don’t need a roll of toilet paper unless you’re truly wild camping—a funeral-sized Kleenex pack will get you out of a bad situation;
• Rope? Come on. I literally brought 90 feet of rope. What the actual fuck did I think I was going to need 90 feet of rope for;
• Jeans are needlessly bulky and heavy—I’d probably go with a lightweight, versatile pant;
• iPad wasn’t needed since I had a laptop and a kindle already;
• I’m still not sure about the hatchet. I only lit one fire, but it was nice to have the option, and there’s something about having one in parts unknown that makes you feel safer.

One thing I found hilarious when packing so tightly was how quickly I came to treasure single serve items like a good napkin stack or zip-lock bag. These can only be bought in household size packs which aren’t so helpful when you just need, like, one.

As for parts and tools, I made the boneheaded assumption that, America being America, I could buy what I needed if, when and where I needed it. Not the case. Not when you’re up against the clock (see previous section).

While I had the basic set of tools, what I ultimately needed and didn’t have was a socket wrench pair large enough for the wheel axle (needed to adjust chain tension), and a spare chain. I will be adding those to my toolkit and bringing spares for all bike-specific, consumable/breakable/bendable parts next time (chain, brake pads, oil filter, air filter, shift lever, brake levers and clutch lever). It also goes without saying that I won’t ever again start a 7,000 km trip on a chain that’s almost EOL. As I said—boneheaded.

Lesson 3: The world is immense

I did some things badly. But I did a lot of things well. I climbed a sketchy hill above a lighthouse to get a proper view of the mighty Oregon coastline. I found my favourite place in the world at the northern trailhead of California’s Lost Coast. I rode the Pacific Highway to Santa Monica and got to see it in all its splendor before the terrible fires took so much from the people there this week. I split lanes on the freeway in Los Angeles and off-roaded in the Arizona desert. I took a 6-hour roundtrip train ride on the oldest operational steam engine in America. I rode on the Million Dollar Highway in Colorado, through a 1 in 150 year Mormon Cricket swarm in Utah, and through hail in Wyoming. I had the time of my life, so far.

Ted Simon, in his 1979 book Jupiter’s Travels, wrote this about motorcycle travel:

“In spite of wars and tourism and pictures by satellite, the world is just the same size it ever was. It is awesome to think how much of it I will never see. It is not a trick to go round these days, you can pay a lot of money and fly round it nonstop in less than forty-eight hours, but to know it, to smell it and feel it between your toes you have to crawl. There is no other way. Not flying, not floating. You have to stay on the ground and swallow the bugs as you go. Then the world is immense.”

Travelling this way, you don’t have the option to grab a coffee for the road. You get it to stay, and you sit with the locals in their favourite cafes, and you take it slow. They come talk to you because everyone is curious where you’re going and how far you’ve come. None of the smells pass you by.

Here’s to the next one, Ted.

Language learning is not something that comes easy to me. My default learning style is to seek out concepts, rules and frameworks, and I’m relatively bad at domains that require a lot of rote memorization and reps. The two that come to mind in that category are language and mathematics, though I’m somewhat passionate about both, in spite of—or perhaps because of—their elusiveness.

I’ve had an on- and off-again relationship with the Spanish language. I took elective classes in High School, set it aside through my 20s, and then in more recent years spent two consecutive winters in Mexico. I leaned on the basic structural knowledge I had during those immersive stints, but my learning was passive. I can get by just fine in a restaurant, but I never seriously set about becoming fluent. I made some measure of progress in expressing myself, but I didn’t have the vocabulary base to reliably pick out words and phrases when spoken at the natural pace of a native speaker.

I’ve changed my approach to learning the language this year, and I think I’ve found two complementary tools that work for me. Neither of them are duolingo. Both are based on spaced repetition, which I now recognize is the way to memorize anything.

From Wikipedia:

Spaced repetition is an evidence-based learning technique that is usually performed with flashcards. Newly introduced and more difficult flashcards are shown more frequently, while older and less difficult flashcards are shown less frequently in order to exploit the psychological spacing effect. The use of spaced repetition has been proven to increase the rate of learning.

Anki

There are all sorts of ways to go about this, ranging from hand-written notes to enshittified SAAS offerings. For my money, there is nothing better than Anki. I will die for Anki.

Anki is a power tool that requires a little bit of elbow grease to become productive with. It offers a lot of settings which can be overwhelming, however it has sane defaults to get you started, and the documentation is excellent.

Here are the two community-made card decks that have become foundational for me:

• Spanish 9000 Sentences contains, well, 9000 spanish phrases along with native audio samples for each. It’s a ‘cloze’ type card, which presents each phrase in Spanish on the front of the card and makes you fill in the missing word. The front might say something like Eso [...] muy bueno. (That would be great.), and the back would say Eso sería muy bueno. along with a spoken audio track. There are 9000 of these ranked and sorted by difficulty. Seriously.
• Ultimate Spanish Conjugation contains full conjugations for 72 of the most important verbs (approximately 59 cards per verb). This one comes with a comprehensive manual for how to use it optimally, which I highly recommend internalizing before getting started.

In addition to these, I also have built out a few decks of my own. Google Translate allows you to export your translation history into a CSV file, which can be imported into Anki. For short, collected phrases like these, I like to use a ‘Type in the answer’ card type to force me to learn the correct spelling, including accents and correct gendering which often can fall aside if you’re just doing pure mental checks in your head.

I’ve also had great success using an LLM to create supplemental cards for knowledge gaps. When you take beginner language classes, you learn all sorts of important vocabularly, like colours, body parts, rooms in the house, telling time, numbers, etc. that tends to get skipped over by the above methods and often as well in real-world immersion. I asked ChatGPT to prepare me several hundred cards containing words and phrases that would typically be covered in beginner or intermediate spanish classes, and it performed very well on that task. This is one of those few genuinely useful applications of that technology.

Pimsleur

Spaced repetition flash cards are great for memorizing, but the other side of the coin is applying that knowledge, by which I mean speaking and listening. I’ve tried in-person classes several times now, and they’ve never worked very well for me. Typically, you’re talking with other beginner speakers who aren’t able to correct you, and who themselves don’t speak the language well, so you aren’t able to train your ear either.

Language exchange meetups are fantastic, but they require a certain level of competency before they become an option.

So how do you practice speaking and listening, short of moving to a Latin American country? For me, nothing beats the old-school Pimsleur Audiobooks. These are also based on spaced repetition, and force you to quickly synthesize responses within hypothetical conversations, closely mimicking real-world experience. Each day you’ll do a 30-minute session where you’ll be speaking, out loud. It’s very effective.

Pimsleur has been around for a long time, and has grown from audiobooks into a more comprehensive app-based offering. I don’t see a lot of value in the other app features, especially as I’m supplementing this with intensive flash card learning. It’s hard to find on their website, but they offer an audio-only subscription (linked above) that is marginally cheaper than the full app experience, and it cuts out a lot of the cruft. There are also other ways to acquire audiobooks, if you know what I mean.

All told, I spend ~30 minutes per day on a Pimsleur audio-lesson, and another 15-20 minutes reviewing flash cards. Combining these methods is a pragmatic, practice-based approach to learning, and is in stark contrast to the academic approach of reading textbooks and completing exercises. Once you embrace it, you may never look back.

Bonus: Cheatsheets

SpanishDictionary.com has an absurdly overpriced subscription tier which gives you access to a collection of excellent PDF-based cheatsheets and reference material. I’ve found the verb conjugation reference valuable as a supplement to the flash cards described above. You can sign up for a free trial, download all the PDF cheat sheets, then cancel.

I’m still here! Made a few updates to the website, including adding (for the first time) a little bit of javascript to lazy load the image gallery posts. I’m still on the fence about javascript as a hard dependency, but I like the look of these long-form galleries and a 35 mb initial payload is downright rude.

In other news, I’ve migrated my primary social media presence from mastodon over to bluesky. A lot of ink has been spilled on this topic, especially in the wake of the election, and I don’t have much new to add. I believe they are two fundamentally different tools, with mastodon best suited for small, niche communities and bluesky a suitable successor to twitter as a general-purpose network. Sensible, top-down moderation of any social network at scale is extremely difficult, and they’ve acknowledged that by building incredible tools for community-first moderation and community-first algorithms. It’s a great community full of kind people and I hope to see you there.