aftd.dev
Web dev, previously a lawyer
1,380 posts
1,098 followers
1,902 following
Regular Contributor
Active Commenter
comment in response to
post
"lengthen and darken my beard" - when your white grandma accepts interracial relationships but not gay ones
comment in response to
post
bsky.app/profile/did:...
comment in response to
post
I'll caveat that a big limiting factor is my ability to context switch. But that was also true at big co job and getting slammed with meetings and PRs and Slack.
comment in response to
post
You're spot on in that intuition figures heavily here as well. Both in knowing what is a good candidate for farming out to an agent and in being able to see an agent's intermediate output and knowing that it's stuck in a dead end.
comment in response to
post
It's rare that an agent is better or faster than what I do by hand. But parallelism makes this about comparative advantage. I get to focus more on "hard" problems while agents handle long tail of "easy" stuff.
In that sense, not too different from being a tech lead with a bunch of jr engs.
comment in response to
post
Not vibe coding per se, but what made agents click for me was having multiple Claude Code instances work on a bunch of small-ish features in parallel.
Ideal case is when there are similar features and lots of boilerplate (e.g. new pref with UI in options page).
comment in response to
post
The year is 2035. You commute to work and strap on your headset. You have an office, but it's just duct tape on the floor and virtual walls. You chat with a holographic projection of your manager, who sits two squares over. You request to work remotely.
"No," he says. "Face time is important."
comment in response to
post
Ah, I get what you're saying. Yeah, this is the larger scale version of "no one knows how to write Assembly anymore"? But worse because there's an added bit of non-determinism.
comment in response to
post
TypeScript is sort of an interesting example in that the benefits seem just as (if not more) useful with AI than humans?
LLMs are bad at context and remembering things. Types are a quick way to catch when they've gone off the rails (ditto unit tests, linting, etc.)
comment in response to
post
We used to be a country.
comment in response to
post
I mean, it's one plane flight, Michael. What could it cost? Two dollars?
comment in response to
post
As someone with bad eyes, I have doubts from a legibility standpoint but I will note that the gradient direction in the Mail app icon (finally!) matches the gradient direction in other apps. So ... OCD folks rejoice.
comment in response to
post
The way updates push content *down* as I'm scrolling is killing me. It'd be great if you detect that I've scrolled and maintain my scroll position or even just did a "5+ new updates" badge or something that didn't shift vertical positioning.
comment in response to
post
I mean, I wouldn't mind more diversity in the sense of "people who don't care about current events and only post about mushrooms" or even "people with strong disagreements about fire safety codes" but that is not what he is talking about.
comment in response to
post
Or a task management system
comment in response to
post
the hong kong protestors used the phrase "do not split" to mean that the radical protest and peaceful protest arms of the movement should refuse to criticize each other.
comment in response to
post
Quisling: The public-private collaboration company. We make the (deportation) planes run on time.
comment in response to
post
You know, Quisling is a nice Silicon Valley sounding sort of name.
comment in response to
post
I don't know much about image hashing, but I think part of the challenge is that you'd want different hashing for, say, photos vs screenshots of text.
comment in response to
post
A bsky / atp feature I'd love to see is the ability for the community to directly annotate an image with alt text and others to upvote. It'd allow friendlier UX than searching the QTs or replies for alt.
comment in response to
post
"Due process for people, not for parking lots"
"Trans rights are human rights. Minimum lot sizes are not."
comment in response to
post
Bezos neutered WaPo for precisely this opportunity.
comment in response to
post
Vivek is in my class. Honestly curious if he'll dare show his face at the next reunion.
comment in response to
post
LAPD is now joining the fray. They are firing rounds of less lethals
comment in response to
post
An audiobook punctuated by random crunches every few seconds, followed by a clatter and "oh fuck", then vacuum cleaner sounds.
comment in response to
post
And 28% approve is ... well ... pretty bad!
comment in response to
post
You could interpret 52% neither + 8% Musk translate into 60% disapproval. That's obviously a difficult extrapolation to make, but 60% disapprove is about the worst Biden ever got.
comment in response to
post
Putin on the phone with Trump: Yes, yes, if the oligarch comes groveling back, you accept his apology but hey, remember when he called the location of his plane assassination coordinates? Anyways, air traffic control these days. Very challenging.
comment in response to
post
I'm logged into Threads and see the same 7 replies, so I don't think you're missing anything
comment in response to
post
Oh wow, where can I get this shirt?
comment in response to
post
I was gonna say Operation Barbarossa makes more sense here, no?
comment in response to
post
I think it'd be very funny if an environmental group used AI to kick off a mass disinfo campaign about mad cow disease to curtail beef consumption.
comment in response to
post
Best burger I ever had was a lamb burger in New Zealand. Still thinking about it eight years later.