Tweet archive

On the other hand, it got very stupid very quickly. It couldn't keep track of several competing requirements: I need to print the output to the terminal AND keep it for later AND have it work over ssh AND i don't have systemd-run available AND setsid didn't do it.

This is the most help ChatGPT has ever given me. I didn't understand select() at all, and being able to ask a patient thing basic questions that would annoy a professional is definitely worth $20/mo.

I should probably clean this up and turn it into a blog post, or at least a gist. Hmm.

(In my program, I do end up handling the SSH case in a semi-ok way elsewhere.)

Was really surprised how hard this was to make work, but I'm really happy that it finally does. If you're reading this and have a critique, would love to hear it.

Comment on what I had to do to get this working OK over ssh here, god terminals are weird: https://t.co/lHKuXNFJjq

I did it https://t.co/erSiNGxqiU

I think I might not be cleaning up my file descriptors properly, although I don't know for sure and I don't know how to check. Ok if you're running on the cli; over SSH or in long running daemon might cause problems.

Ever wanted to save stdout/stderr for processing in a Python program AND have it print to the terminal in real time? ME TOO.

Copilot trying to complete my sentences when writing a comment is super distracting. Realized today that swearing shuts it up. # Do X because of Y becomes # Fucking do X because of Y Also now my code is more fun to read. :)

@SwiftOnSecurity Oh yeah we should definitely be taxing use of XML

All people, RSS is 24 years old. Maybe it's time we get browser support for RSS links going to dedicated RSS consumer software (whether that is an app separate from the browser or a website accessed by the same browser)

Feed Hawk was great. The replacement in Unread is good if you use that. Various readers have some implementation. But fundamentally a better approach would work like mailto: or tel: links, with an OS preference for preferred app.

RT @nikitonsky: I wonder why no app has figured rich text editing yet. It’s really simple: you allow two positions on block boundary, one i…

I wonder why no app has figured rich text editing yet. It’s really simple: you allow two positions on block boundary, one inside and one outside. https://t.co/85RFaoRGAA

@Jowjoso Oh I see, didn't realize there was a paid tier for the Reddit one. Would actually be interesting if you could monetize a large social network without ads at all, as a service for ai training. Monetization wouldn't require high in app times, so maybe less dark patterns too. Hmm

Me: "I ALREADY VALIDATED THE XML IS CORRECT!!!" Computer: "Sin tax error please enter your credit card"

@mrled it's important to note that the api is not actually deprecated, just moving to a subset of data or requiring pay, depending on which company we talk about. you're also right of course but other incentives, I just hadn't considered this arg before, and it probably is A Factor?

@Jowjoso I dunno about this, seems like an api that requires registering and comes with a eula would be a more likely way to enforce a rule against downloading too much. My guess on both counts is that the company values time in app and doesn't value the content otherwise.

I have been thinking about why this was happening all at once, everywhere, a lot and hadn't yet thought of the "it's a training most" argument oh. that's probably true. at least to a certain extent. that's so fucking dumb I hope full scraping breaks everyone's service. https://t.co/r6vrON0Hqo

upon reflection, after seeing the X corp stuff and apparently reddit is killing its api too, I think the "bots" excuse for the twitter api was bullshit and anyone with data is going to use it to train a shitty llm while trying to block real ai companies from gathering it

Some people need reverse lmgtfy. Why Did You Drop That Link In The Chat And Them Fuck Off For Seventy Two Hours. https://t.co/AADTRiobg2

GitHub should have a cron job grep for "I don't know how to open a pull request" across all GitHub issues and just delete the account of anyone that posts that text. Why would we even want you here. https://t.co/C16DS8sawK

"Oh that thing I had nothing to do with belongs to me, give it to me by force of law" AI is so cool. We can all sound like Drake now. https://t.co/w4XoOTIFzD

Stories about working at Amazon are kind of hit or miss. Definitely some cool work to do there, but some things just sound like weird own-goals. However. The written culture thing is objectively correct and I would LOVE it.

Almost nothing makes me as mad as emails like this https://t.co/7OxAPCy2ZZ

brb sending a 4 paragraph email to my boss with questions about a 20 person meeting he included me on during a time I was marked as OOO whose title is keyword soup and whose only agenda is a spreadsheet with 8 tabs

A long time ago in my career I was trained out of sending multi paragraph emails, as no one reads them. I still have the urge to do this and I still think I'm right to do it. It's better than telephone, Slack, Jira, or Google Calendar, and I gave all of those a shot.

(I don't like that I frequently have to refer to declined events, but at a big company that does things by no-context calendar invites, the agenda field of an event might be the easiest way to find a document you need later, even if you didn't attend the meeting.)

It shows the same event on multiple calendars as just one event. I love this bc I can invite my work email to events from my personal email to properly set work availability without clutter.

Tasks from Reminders on macOS/iOS show up at the top of every day after their due date until they are complete. This means tasks don't get lost and you also don't have a notification that stays around forever.

It is sadly missing a toggle for "Show declined events" on the main calendar page, however, on macOS it does have a menu item for this, and you can add a keyboard shortcut in macOS Settings -> Keyboard -> Keyboard Shortcuts... -> App Shortcuts

Previous context: https://t.co/6ZI6KM2bbF

Natural language calendar input really is great. "Dinner at 530 on Sunday with Hannah at Geraldine's" does the thing you expect.

It doesn't send me a push notification every time someone at work adds something to my calendar? Why tf does iOS do this.

Thread for my favorite @flexibits Fantastical features, so that I can remember what it helped me with

Imo, our copyright system should prioritize access to old cultural artifacts. A simple change: if something is not made easily available to the public, it should lose protection. If it's not making anyone money anyway, the cost would be small. https://t.co/Up5xr2fyf1

This is on my mind bc I think it's frustrating when being data driven is held up as a goal with no downsides. The upside is a certain kind of pride in your work - but the downside is the loss of a different kind of pride in your work. You cannot have them both.

Also - and this is just my subjective taste - a design-driven project is much more interesting to work on than a data-driven one.

(I don't mean a front end design -driven project thats about writing CSS, I mean any project with a user in mind and opinions on the best way for that user to get work done.)

Also, good design must be made by individuals, not groups; the best designs will come from organizations where a single person can make an executive decision to overrule everyone else. (This is also where the worst designs will come from.)

By contrast, design by committee can achieve some floor that is better than the worst possible designs. (Less strongly held.)

It makes me honestly worry for some things I do love. What does the data say that gists do for the stock price? Probably not much! If it ever breaks, will it be worth engineering time to fix? https://t.co/TX5GS6wy75

Another nice thing about being data driven is that it is more compatible with distributed decision making. 8 people in 4 time zones can look at a graph; twins who have never been apart might not agree on the most beautiful color scheme.

I think one of my low level beliefs is that good design requires taste, and since taste cannot appeal to everyone, a good design must exclude or repel some people. There is no such thing as good design for everyone.

Another Microsoft example: https://t.co/p9CMHGhvv0

An example: this is obviously a bad experience for anyone who uses it, and absolutely tacky. It feels cheap, frustrating, like a used car salesman handing out business cards at a child's birthday party. But I would bet serious money that it makes $MSFT 📈 https://t.co/SlqyYvGBXt

Meta has done this forever. One nice thing about this is that you don't need a theory for how to make money in the medium/long term, just run short term experiments. "But it truly feels like Instagram is Gromit laying tracks just ahead of a fast-moving train". Indeed. https://t.co/3dJnwiivdu

Another Instagram example. The head of ig knows people don't like it but will use it anyway. It's not even a question which one they'll choose. https://t.co/0X2SnbHX3o

You can build an algorithm, relentlessly optimizing for revenue (or some proxy like time in app or whatever), and this can make you a lot of money. However, past a certain point relatively early on in your products life, this will make it subjectively worse for you average user