The argument made by some folks regarding LLMs being non-deterministic is that humans are non-deterministic as well. However, humans are non-deterministic in a particularly human way. Also, humans can (usually) explain themselves (at least in terms of why they chose some particular solution), but LLMs cannot explain their output (despite appearances), since it’s stochastic, with no underlying meaning.

PSA: The Amazon wishlist doxing threat is much greater and more immediate than folks might realize. Attack works like this:

Stalker who wants your address opens an Amazon seller account and lists themselves as a third party seller for any item on your public wishlist. Then, they order the item from themselves as a gift for you. Bam, they have your address.

In particular, attack does not depend on an existing third party seller having poor PII handling hygiene, like the articles have implied.

It never ceases to amaze me that people really think an LLM-based "AI agent" will deterministically heed their instructions.

Because they believe it "understands" them.

When I use a computer, I don't want to rely on luck. And I never will. Leave me behind, thanks.

Prospect Theory shows that our voice calculus tends to shift towards the negative.

https://psychsafety.com/prospect-theory-and-psychological-safety/

Interesting how quickly everyone is now becoming a lifelong TDD practitioner now it appears essential for GenAI assisted development. The reality is that only a tiny fraction of the industry have genuinely been practicing TDD in the 20 plus years since Kent Beck first coined it.

I'm low key impressed that the AI corps have found a business model:

Data Privacy as a Service

Don't want your data to keep feeding their models which their shitty security means others have the chance to extract? Pay up.

I've just published a reference for anyone using git while trans, explaining how to hide or remove your deadname in git repositories

Please share ❤️

https://code.curly.kiwi/2026/02/27/using-git-while-trans/

I believe that the #accessibility of everyday tech for #screenReader users is on a slow but consistent decline. Operating systems, browsers, messaging apps, email clients, even command line tools.

These things are not being replaced with more #accessible alternatives, but nor does the investment exist to stop the rot within the current options.

This in itself is concerning, particularly as it mirrors tech trends more broadly. But what I worry about quite a bit is what it does for user expectations.

What happens when generations of people grow up with inefficient keyboard access models, faux desktop apps, and a thousand tiny papercuts here and there? When the new baseline is worse than it was before, it takes that bit more effort to imagine and advocate for best rather than just better.

CEOs who genuinely believe AI is capable of replacing this many people do not understand what their workers do, or what AI is able to do.

None of this is because of AI gains, it’s either deep incompetence or incredible greed from execs and shareholders.

Meanwhile the platform will get worse.

How can I explain to my boss, that it is a terrible idea, to install OpenClaw in any environment at work?
Explain it like the person does not know how computers work.

#infosec

It's one thing to say things are bad while offering suggestions/doing the work to make changes.

It's another to just hand wave about it being "doomed" while doing nothing but complain.

If you're the latter, let me just say the former is way more fun.

Bonus: I'll find you way less annoying too.

AI, in the way it's used now, is automation but without control, without predictable outcomes, and without transparency.

We can have automation with control, predictable outcomes, and transparency by the way. We've been doing it for decades before. What if we put money in that instead?

#NoAI

"The boys’ club: How Epstein’s influence shaped the exclusion of women in STEM" by Jessica Kutz: https://19thnews.org/2026/02/epstein-files-academic-research-women-scientists/

Nürnberg bleibt stabil!

Heute sind wir mal nicht dem Team rechter Menschen (der Haufen schrumpft auch jedes mal mehr, Gegenwehr hilft halt doch) hinterher gezogen. Dafür haben sich drei Gegendemons an der Lorenzkirche zum Rave gegen Rechts getroffen.

#nürnberg #ravegegenrechts

Wer sich noch nicht entschieden hat, welcher Samstags-Gottesdienst besucht wird:

Natürlich findet auch heute ein Vermummungs-Gottesdienst statt.

Nach dem Rave gegen Rechts am Westportal der Lorenzkirche. Ca. 17 Uhr.

[Ausnahmsweise nicht bei der Abschlusskundgebung des TrM (Jakobsplatz)].

@inw die Alerta-Parade verlässt vermutlich am Plärrer die TrM-Strecke und läuft durch die Innenstadt zum Rave gegen Rechts an der Lorenzkirche.

Technical coaches help developers around them to adopt more effective engineering practices. I like to lead team 'learning hour' sessions to teach specific techniques and have fun discussions. Training coming up this week, places still available: sammancoaching.org/training/del...

Deliver Learning Hours - Samma...

When exactly did we start to think that the point of humans was to be productive, instead of thinking that the point of productivity was to take care of humans?

“Terry Pratchett wrote every decent person’s favorite summary of the bedrock of humanist morality in a fantasy book for young readers called I Shall Wear Midnight: ‘Evil begins when you begin to treat people as things.’ Treating people as things can begin in many ways, but I think one of them is the idea that things can be people.” https://www.todayintabs.com/p/a-i-isn-t-people