"Books as signifiers, the paradox of tolerance, and Nazi bars"
www.baldurbjarnason.com/2026/paradox...
On how books work as social signals.
And a cursed book.
Posts by Michael Laccetti
In this case, MFA via TOTP, not email. If an attacker has access to my email, and it is used for MFA, I'm boned.
I'm curious about the security implementation that requires emailing a code to me, THEN using TOTP for MFA. Why not skip emailing a one-time code?
Am I misunderstanding how MFA works? If I have it configured on my account, why are you emailing me a code, only to then require my MFA code? What's the security posture, here? #security #mailgun
Apple is now two for two:
* Camera bump when Google did it was terrible, but when Apple does it, then it is awesome
* Glass effect was terrible when Microsoft did it, but when Apple does it, it is a delight
Colour me unimpressed.
@edzitron.com Reading "Make Fun Of Them" and laughing about the bit where Google is juicing their MAU numbers by ramming Gemini everywhere.
Whether you like it or not, we rammed AI down your throats, and now we're monetising it by jacking up prices. Then, we get to say that our AI-based revenue is up 350% to justify all of the data centres we've built out, and we also get to say that our DAUs and MAUs for Gemini are way up. Nice.
2. If this really is an agent, why do I need it to lock up my editor while I'm working on it? Edit files in the background, or leverage Live Share - pretend to be a pair programmer with me. Be autonomous, and complete your task(s). Cursor seems to be lazy, and will collide with my work.
In the continuing thoughts (rants) on AI-powered coding (see: vibes), I offer up two thoughts:
1. This needs to get faster. Way faster. If I'm going to burn the same number of API requests, work on multiple files at a time. Parallelize the crap out of it.
I've found myself thinking that I would have been better off writing it myself. That I would have had more context on what is going on, and how to fix it. It's an unpleasant experience to outsource your thinking and realise that was the wrong decision.
Also, having to watch it do it wrong in many different ways. If this is an "agent" - we can do better.
I'm not saying that I don't find it helpful, just that it feels very surface level. The second you go deeper, it just exposes a lot of the gaps/issues.
Knowing that I'm paying, and watching it try and retry a few times before it gives up, picks another random approach, and starts all over again... 👌🏼 Loving it.
It feels like watching a co-op student (intern for those of you who didn't go to UW) search the internet, pick a random solution, and then try to kludge it into working order. Except it doesn't.
I also am a _HUGE_ fan of when it randomly pivots and decides that it must import random items, or generate code completely unrelated to the task/request. That's just lovely.
Are you using something new that it hasn't been trained on? Expect a fun ride of chaos and random decisions! Want to use v3 of something, but v2 is more popular? Enjoy the mish-mash of incorrect code!
It frequently loses context, forgets instructions, or randomly picks something from LLM-land. If training data is from GitHub and Stack Overflow, then it'll be creating code the follows suit.
Continuing with the AI (Cursor) thoughts:
I become incredibly irritated and/or angry with Cursor and the code generation. I can't tell if this is due to me spending more time with it or changes made upstream.
Is it incredibly useful for building some scaffolding for UI stuff that I am not familiar with/good at/excited about? Totally. The implementation of the logic, though - that's where it falls down.
Hot take: vibe coding is only suitable for generating revenue for AI developers because you have to prompt and re-prompt your way out of the complexity and bad decisions that your codebase becomes. Each of those agent-based prompts? It isn't free.
I finally got my opinion piece out on Toronto Driving (laccetti.com/on-toronto-d...). Lack of homogeneity, combined with no enforcement, is our biggest challenge.
Teslass are the new BMW drivers, anyway.
A snippet from the Globe and Mail article about Oliva Chow and what she's doing to reverse Toronto's decline by getting people to come to the office four or five days a week.
I was reading an article in the Globe and Mail about Olivia Chow's efforts to revive Toronto's decline (www.theglobeandmail.com/business/rob... ). It is rare for a politician to acknowledge that things are tough and that they're moving levers. Less sold on the "force people to the office" idea.
Take a lane from Lakeshore and Queensway in each direction, and throw in more trams. That'd hugely open up a bunch of density/liveable areas in Etobicoke/Mississauga that are criminally disconnected from transit.
Great that our money is being spent on spreading the good word. Nuclear is 100% "clean" if you ignore any byproducts/long-term storage of the waste. And highways? We don't need this 413 nonsense - build more public transit, please. And actually deliver it - that Eglinton Crosstown is a disaster.
I'm driving my daughter and her friend to school, and they insist on listening to the radio (Gen Alpha wut). This is the first time in a while that I've heard advertisements, and some good Gov't of Ontario propaganda comes on—talking about "clean nuclear" and "building highways."
This is amazing.
Instagram Reels: softcore porn
YouTube Shorts: every right wing talking point with weird Christian shit thrown in (seriously - Ben Shapiro and Jordan Peterson, wtf?!)
These algorithms really know me.
Forking nginx (mailman.nginx.org/pipermail/ng...) is a strong statement about how F5 has stepped in to "guide" the project.
Anybody else wonder if the designers had the wrong takeaway from the Terminator movies?
electrek.co/2024/01/12/h...
The humidifier water that costs $20k and needs 6000W to generate. Better wire that up to your own wind turbine.
Cue the hype train.