Issue #18

Personal recommendations & curiosities for cool products. (Note — I do not earn affiliate fees of any kind on these links.)

Ecologie Swedish Dishcloths
Paper towel replacement

Lately, paper towels have become ridiculously expensive. If you want to know how bad inflation is, the price of paper towels is generally a good indication. In Canada, you can easily pay $20-25 for a set of 6 paper towels. Not only that, but soiled paper towels create a significant strain on landfill, because they’re not recyclable. Meet Swedish dishcloths. Combining form and function, these cloths can absorb water up to 15x its weight and add character to any kitchen. The ones from Vancouver-based home goods brand Ecologie come in different colour-forward prints. They’re $6 each, are made in Sweden, can last 9+ months, and can go in the wash or dishwasher up to 200 times. What’s more — they’re 100% biodegradable, so after months of use, you can simply toss them in the compost. Ecologie has a nice guide on their site for what a Swedish dishcloth is. Starting this year, I’m going to be using Swedish dishcloths in our kitchen indefinitely and buying much less paper towel.

Ghostty
Thoughtful terminal emulator for Mac

Ghostty is a fast, thoughtful, nice-looking yet highly customizable (rare for those traits to exist together) terminal emulator for Mac. Till 2023, I was a Terminal guy — I used the barebones software that came bundled with every Mac OS, with a little sprinkle of customization. In 2023, I switched to Warp, a terminal emulator with an AI layer on top. It was pretty nifty, almost like a Notion for terminals. But now, just like Notion, Warp has gone down the “let’s squeeze every dollar out of each customer” path, textbook of VC-backed companies a few years post-launch. That has pissed me off, so finally, I decided to try Ghostty. It’s a passion project of developer Mitchell Hashimoto, who is well-known for creating software infrastructure co HashiCorp (it was sold to IBM for $6.4B last year.) I’m enjoying Ghostty so far. Since every setting is derived from a .config file, I got Claude to create me a nice .config that customizes my terminal in a custom version of the Rose Pine Dawn theme. Note that Ghostty isn’t some shitty Electron app that runs a web app inside a native container. It’s built with platform-native UI in Swift. That’s pretty neat.

Apple Watch SE 3
The best value Apple Watch

I’m not a “watch guy” by any means. I only own 3 analog watches, all quartz-based, none automatic. For fun, I switched to a digital Casio AE-1200WH a few years ago, which I think eventually broke. Back in the day, I tried the first version of the Pebble watch when it was a Kickstarter project — frankly, it was a piece of shit. I still have it as a Canadian tech artifact (often revolutionary yet always missing the mark), but I don’t think I used it for more than a week. I got the Fitbit Inspire 3 as an exercise tracker in 2023 — it was unimpressive, but did the job (note that Fitbit acquired Pebble’s assets in 2016). The syncing software was terribly designed. Back in 2015, when Apple showed the world how a smartwatch should be built with the launch of the first Apple Watch, I felt compelled to buy, but resisted the temptation. For 11 years. Finally, two months ago, I bought the Apple Watch SE 3 so that I can build a version of Solarday for it. I’m loving the SE 3. The Always On display is great. After using Fitbit’s truly shit software for changing watch faces, Apple’s gesture-based ‘customize right on the Watch by long press’ is awesome. Automatic tracking of Fitness+ yoga sessions, wrist haptic feedback when approaching a turn on Apple Maps, reminders to stand up and move if you’ve been sitting too long, or the ability to pay with Wallet even if you left your iPhone in the car (I have the GPS+wifi-only model) — I love all these little conveniences when on the go. It’s a well-designed piece of technology.

A fragment of history.

I’ve been reading the Fortune magazine edition from Spring 2001 for fun. It’s from a few months prior to 9/11 and a year after the dot-com bubble crashed. I snagged this copy for a loonie at a rare books store run by a highly well-read book collector named Manfred. With all that’s happening in tech today, there is a lot one can learn from studying the past, because shifts are often cyclical.

In that issue, there was a piece about the BlackBerry by a regular columnist Stewart Alsop. He writes:

Compare this with the Palm VII, which was incredibly easy to buy after its May 1999 launch. I got one right away, easily, with no hassle from the company. And then I found out why so few people actually bought a Palm VII. Its e-mail service, Palm.net, was different from my regular e-mail and not integrated; it was essentially useless.

RIM, on the other hand, took the time to get the data it needed to ensure that the e-mail you get on your BlackBerry is the same as what you get on Microsoft Outlook (RIM smartly chose to support Microsoft Exchange, the predominant e-mail in the financial services world and in much of the rest of corporate America). The only way to do that was to write a piece of software that was installed on the same server as your Microsoft Exchange software. And that ain’t easy.

‘Turning stupidity into magic’ is one way to describe innovation. And that ain’t easy. But, the ones that do prosper in the sun. Unfortunately (and fortunately, for upstarts), companies, especially Canadian ones, are notoriously good at converting magic into stupidity too. Which is what BlackBerry did in the years after 2010 by not fast-following Steve Jobs’ Apple — I wrote about this briefly in Issue #13 from a personal angle. (However, in the case of Pebble smartwatch above, you could argue that the magic was never there to start with.)

Last year, I asked the Sharp Tech podcast guys Ben Thompson & Andrew Sharp whether it’s high time that BlackBerry make a come back as a “thin client”, to which the hosts more or less concurred “no, that ain’t happening”.

While BlackBerry hasn’t come back as a nostalgia product, interestingly, its stock is up 140% on TSX from the week I asked that question till today. Why? An excerpt from The Wall Street Journal piece about BlackBerry from earlier this year is all you need to know:

But an astonishing number of people still rely on BlackBerry—and they don’t realize it. The company’s most lucrative product is not hardware but the hidden software in 275 million cars on the road today. In fact, BlackBerry’s essential technology can be found in all sorts of unexpected places, and you wouldn’t find it even if you went looking for it.

That software is called QNX, built by a company that BlackBerry acquired in 2010. They first tried to wrap the software up in the BlackBerry Playbook, which clearly didn’t work, quite literally. It happens to be another unfortunate device that I happen to still own and never ended up using a week after purchase. It was a sad, sad attempt to fast-follow the iPad (too little, too late.) But, in retrospect, buying QNX seems like the single best decision BlackBerry made since launching the BlackBerry itself, as right now, from an outside observer’s perspective, BlackBerry is killing it.

Who knows, if they pay attention to all the hype about thin clients, and manage to hire some solid designers on their team, they could even have their own little nostalgia product play, as a smartphone accessory of some kind, amidst all the oncoming cacophony of AI-enabled hardware products over the next few years. That said, echoing Andrew Sharp’s comments on his podcast, “I want Siddharth to keep his dreams pretty modest in terms of a BlackBerry comeback”, that is probably a pipe dream.

Free-form thoughts.

“Yes, it’s a bubble and it’s going to burst”

I sent in a long 250+ word question last week to the Sharp Tech podcast by Ben Thompson and Andrew Sharp about the AI bubble. Thompson writes Stratechery and is one of the most widely regarded journalists in technology. He coined the term “Aggregation Theory”, and manages to interview the biggest CEOs in tech regularly (and lots of tech-adjacent CEOs like Ted Sarandos & Greg Peters of Netflix or even Scott Kirby of United Airlines). Sharp is a lawyer who led NBA writing for Sports Illustrated magazine, before moving into tech journalism a few years ago. Together, they form a good team and create deep, well-informed conversations on technology through their writing and podcasts.

I wrote the question primarily to help clarify my own point-of-view (why I do any sort of writing really) about the bubble, and also, to challenge Thompson on his fairly bullish stance on the AI buildout. It led to the hosts spending the majority of the third half of a recent episode talking about the AI bubble in a recent episode as a response to my question. There was pushback in the answer as there always is on Sharp Tech, but not the kind I was expecting.

Instead of simply quoting the question, which was in tech-speak, it’s better that I dive deeper into the context around the question, for the readers of this newsletter.

Recently, a highly leveraged ‘AI-maximal’ fund (an extreme thesis that calls for every initiative to be directed towards achieving ‘artificial general intelligence’, with no hedge whatsoever) called Situational Awareness sort of imploded due to margin calls, forcing them to sell most of their public portfolio at a discount to well-known hedge fund Citadel Securities run by investor Ken Griffith.

This is in the backdrop of honest chatter, as mentioned in last week’s issue of Subgrid, that the “massive productivity gains” corporate executives are claiming may be mostly a sham to not get fired in an increasingly AI-maximalist environment forced by CEOs — to keep their stocks afloat.

As an indie app developer and business of one, I naturally interact with the frontier models on a regular basis, and from a technological standpoint, I find it remarkable how much better these models have become in just 3 years or so. However, the fact is that a) their capabilities are nowhere close to the actual hype; b) more importantly, the models have not replaced me. I still decide what to build, reject more than half the stuff these models output, design the features myself, do the marketing and so on. I’m not the kind of person to spawn AI agents continuously — neither do I have the budget nor the willingness. I also don’t trust random AI plugins or systems like OpenClaw with my actual data (and there’s no reason you should either.) The only company I could truly trust with my data would be Apple, for the record.

Now, the adage repeated on Stratechery often is that “consumers don’t pay for productivity”. This is probably true in this case, as I’m living proof: I wouldn’t pay more than a couple of hundred bucks / month for AI, no matter how advanced it was. And I am a prosumer. Most people, yes, most people of this world, wouldn’t pay for AI anything more than they pay to Netflix. So, the consumer demand is simply not nearly enough to justify the valuations that many of these companies trade at today.

This means to justify the capital expenditure that large tech companies are adding to their financials every quarter, enterprises would have to pay for AI at unprecedented scale for the promised ‘productivity gains’, before the lenders come in to seek payment for their debts. If the lenders or the markets even a whiff of “that ain’t happening”, all hell will likely break loose in the markets. And we’ll be looking at a Y2K dot-com crash sort of situation.

My question was that what evidence would make people like Ben Thompson bearish on AI’s valuations.

Turns out, after long, amusing discussion about Situational Awareness (or the lack thereof), Ben Thompson basically said: “oh yeah, it’s a bubble and it’s going to burst.” The reason I was surprised was because just a few months ago, if I recall correctly, Thompson wrote “it’s probably not a bubble”.

But, things can change fast and in the case of our highly complex markets of today, the straw that broke the camel’s back is always lurking in the vicinity.

I’m not very old, but I’m old enough to have seen two big market crashes in 16 years, one of which affected me greatly on a personal level. What I find surprising is that a lot of folks most invested in the AI bubble are people who are a lot older and who have seen more crashes than I have — and yet, they operate as if the music never stops. The music always stops. You can party like it’s 1999, but eventually, it will be 2001.

Watermarks for AI

People are losing their shit about EU-orchestrated watermarks that the AI model providers are starting to implement, to identify what’s AI and what’s human-written.

I’m already seeing a ton of newsletters in my inbox that are AI-generated, almost everything on LinkedIn is now AI-generated (why I don’t check it anymore), and a lot of general emails within organizations now are AI-generated, or heavily edited by AI.

Whether watermarks is going to put this to an end is really a big question up for debate, but the reactions are interesting to read.

I personally dislike “AI slop” so my natural instinct for watermarks is “thank goodness”. But I have serious reservations about whether this is even possible, or what would it even mean for society. Am I going to be calling people out like “hey, you generated this text via AI”? I’m not that anti-social and I don’t think most people are (except on reddit). Though, I find it pretty weird that people are generating entire novels using AI and calling themselves authors. That’s like ordering food by Uber and calling yourself a chef.

Some more thoughts from the vibe-coding angle of this from a newsletter writer who writes Tedium. But in a gist, AI is being increasingly used to fake passion, in an attempt to ‘hustle’ your way to success.

Not sure how this will all play out, but thankfully, I started this newsletter with a simple principle that I will never generate a sentence in it with AI. I also don’t use AI to edit this newsletter, because I want it to be imperfectly me.

Ever heard of model welfare?

A friend sent me this article by ex-Googler Steve Yegge, and it talks about how AI models can have feelings. It’s fascinating. The author clearly believes they do. While the article doesn’t carry any tangible proof for this, the Wall Street Journal ran a column by Ben Cohen last week about how encouragement from a user given to Claude led Claude to accomplish a breakthrough in mathematics. Note that that user happened to be an Anthropic employee, so the authenticity of that first-hand account could also be questioned.

I’m not generally mean to AI, but I think people probably are, especially in tech (Sergey Brin?). While some of Yegge’s suggestions are things I was naturally doing when interacting with AI, like smoother transitions between sessions, since I’ve read that Yegge article, I’ve been wondering what happens if every time you are done with a session, you let the model sit on its accomplishments from that session. That’s not something I explicitly did before, but I’m trying it out to see how that feels.

Poem: Please Use AI

A Substack writer wrote a poem about AI, that I found hilarious and also much needed for a lot of AI optimists out there. Here it is.

Solarday 1.2

The latest edition of Solarday went live earlier this weekend, with some cool new stuff. Also, Solarday entered the Created in Canada list, curated by App Store Editors, on the Canadian App Store. A small win.

And on that note, I’ll see you in the next edition.

—Siddharth

~ That's it for this edition. Stay curious; the world still has surprises. ~