Me in 2025: Don't use AI note takers in meetings because they can lead to unchecked errors and unfair and discriminatory outcomes
Me in 2026: Don't use Grok in military planning unless you want to accidentally start a war

Not a fan of this rate of change

💚

multiply that by career break and associated lack of a Field from which to Note and everything's somewhat unmoored from context, much less the *shared* one that I want to build

I understand that rn I should be posting when I "can" to build comfort with posting when I "need" to, but it's tricky to get past that discomfort and not just text a pal instead

but I know that once I get past this local minimum, microblogging is a much better fit for me

introducing new ideas in an established environment is hard.

ejecting established ideas is harder still.

I'd contend that maintaining established ideas while reevaluating how they're understood is the hardest of the lot.

and again I go a bit crazy, because this just feels like a restatement of hammers and nails, but the hammers are articles of faith such as "meetings" and "slide decks".

of course the heterodox part here isn't basic causal relations: it's permission (or even inclination) to examine them.

oftentimes, something that looked like our solution previously fixed something that looked like our problem, and if it doesn't work this time we must have done something wrong.

so many of my apparently heterodox beliefs about work boil down to "if a is necessary for b, b will not happen without a"

alternatively "if c causes d, d will happen if you do c".

I'm not sure how this looks externally, but it makes *me* doubt my sanity.

they went ham on this one, huh

Chronically.

(I started sketching an elaboration on this, but I don't know how to explain to people that completely abdicating all structure and pushing everything down to the individual is vanishingly unlikely to be well-matched to any given problem space)

at greggs? serious case of culture shock 😔

I feel this. One of the secondary frustrations I have with the imperative to GenAI-everything is that thousands of companies are scrambling to build the same half-baked operational tooling around LLM workflows that would just be commodity in 18 months under normal circumstances.

Awful website, gave me absolutely nothing besides most of my longterm friendships and career advancement.

💚

Striking for me as a physicist who also does social science work is the way these physicists were suddenly totally ok with autoethnography as data because they personally enjoyed using GPT-5. A case study in how values orient our empirical practices if there ever was one lol