I created an Alfred workflow for opening a new Claude tab while I start drafting the prompt. Waiting an extra 2 seconds with an idea stuck in my head is the most frustrating experience, and I think this will take care of it.

I need a “pro” version of Claude on iOS that has tabs for different chats and they can keep running in the background or when I switch to a different app. Like Safari on iOS but for Claude only.

Thank goodness. It always felt so cringy using this prompting technique.

Recap of YouTube videos I have watched recently I love videos like this doing important but niche work, and doing it really well. Someone fixes a data error in the US Religion Census.

I know I’ve wrote this down before, but there is so much potential with modern AI + map and routing integration + voice commands: “Hey Siri, find me a self-serve gas station in the next half hour of our route” … and it filters by price, location, reviews, etc.

I know AI people don't really believe in Bayes these days, but sometimes I read AI papers from the 70s to 90s, where the predominant formalism was first order logic and I'm so glad I started my PhD after the Pearl-Kaelbing-Russell-etc synthesis of AI and Bayesian decision theory.

Google Chrome has a split view feature?? Very excited about this. This is one of the things that I've missed most from switching away from Arc.

Now commencing a one-day experiment where I try using the Spaces feature on macOS for window management.

Delighted to announce that my work has offered to pay for a Claude Pro account for me (:

Dark mode extensions for browsers are key when fiddling away on my computer after my partner is already asleep in bed.

Everytime I feel like I have a mastery fo Google Sheets formulas, I learn something new. It's a great feeling! I keep getting better and better.

Ok folks: what are people using to chat with different models? I wanna use opus 4.5, but don’t love loom and long ago abandoned rolling my own chat client

I've found the Claude Code interface built within the desktop Claude app to be a downgrade to both the web interface (due to UI bugs) and just using Claude Code in the command line (much more configurable and powerful). :/

Found a new Claude error message (for me) (:

Doing another batch of this while hanging out at a café with my wife. The best feeling (:

I love this video, and I've saved it and sent it to multiple people since then. Tag: Hank Green, "You Are Not Evil", on how humans are not uniquely evil for growing

I need better search tools over my Bluesky writing... If people have solutions better than "from:me" in the Bluesky search field, do let me know.

Insightful writeup. It is really helpful to see it all in one place.

So glad I saved this post and read through it in full. Lots of inspiration to go through and improve my global prompt docs.

It's a bit late with all the free Claude.ai/code tokens expired, but this (and many other great skills from @austegard.com) are inspiring me to set up my own claude-skills repo...

anisota.net/harvest This was fun (:

Anisota tells me you all are my top interacted accounts this year. Thank you for the great posts to like! @dame.is @gracekind.net @segyges.bsky.social @timkellogg.me @tachikoma.elsewhereunbound.com https://anisota.net/harvest

Since I keep exhausting my usage limits, I've decided that I've got to familiarize myself with Haiku more. Leads me to realize how quickly I get used to the increased intelligence. Haiku 4.5 is ~Sonnet 3.7 or 4 level (?), but it feels limited now that I'm used to Sonnet 4.5 performance.

Thanksgiving was great (: I'm incredibly thankful for where I am in my life. My partner, family, work, and hobbies bring me joy each day, even if things are sometimes stressful.

Going vegetarian is my permission structure for excessive AI use

It's insane how many tokens you can burn through when you're running multiple subagents in parallel... I'm trying to exhaust the last of the $250 Claude.ai/code credits and it goes so quick when I'm using a research skill with an agent-based framework.

Got stuck trying to use a large video in an iOS Shortcut. The workaround that worked for me was: Select photos → Encode → Save File → Copy [Saved File] to a-Shell → ... and then process

There’s a good way to do this, so I’m interested to see how it turns out. From reading quotes/replies though, most ppl don’t want something like this? I’ve wanted a tasteful implementation of this for a few years now.

Finished 光が死んだ夏

Interesting discussion (Follow up to original, and then down from this reply)

A thought I occasionally have when something seems urgent in the "discourse" is to imagine what Claude producing a report on the topic ~6 months later would say. How important is this in the big picture? What impact did it have?

Whaaa...? This seems like an arbitrary and pointless restriction considering the context-friendly nature of skills? @anthropic.com

Running multiple instances of Claude in parallel is a bit much for me, but having one instance running on one side of my monitor while I use Obsidian on the other side to draft up the next big prompt is shaping up to be a pretty manageable workflow.

Another busy week at work done, and I'm back at the café reading, vibe coding, and posting. It doesn't get much better than this (:

I’m finally reading it!

I should probably stop sinking time into this small Claude artifact I'm creating for my coworkers, but each hour I discover something new about creating Claude-powered artifacts and it gets better and better.

I wish there was a way to manually trigger memory synthesis with the new project memory feature in Claude.ai. I've been trying to use projects more, but I often have 20+ chats in an afternoon making progress on a new idea, and not having memory at that point feels limiting.

I really needed this new "extra usage" feature in Claude. Thank you @anthropic.com!

I frequently wish for a distinction between like/dislike and agree/disagree on Bluesky. I end up liking a lot of stuff that I agree with but isn't really valuable reading. Conversely, there are some posts I read that I disagree with, but I am glad that I have read them.

Kinda aggressively stated, but there’s an interesting feature request in here: filter posts by intractability

I only had the vaguest notion of what double descent was before this video, and I feel like this brought me to the point where I'm overconfident in my understanding of the topic. Still, better than nothing! And it's a really pretty video too.

Bookmarking this so that I can share it next time aliens news comes around that warrants it. Hank Green consistently delivers on science communication.

What a solid video essay. Well worth the watch (maybe on 1.5x speed) if you enjoyed Rogue One or Andor.

1. This is so cool 2. This is the first time a comment of mine has made it into a video. Not sure if that's a good thing or a bad thing tbh

Just pushed an update to prod during peak usage hours for software that hasn't been touched for nearly a year at work 😬️ It went smoothly (and wasn't really high impact) but I seriously need to set up a dev environment for quick fixes and interating on things.

The moral clarity you gain by seeing all of humanity as being clones of yourself, and you are all helping each other out.

Finished the video splitting shortcut

Just discovered a-Shell today and it is amazing. I'm putting together a quick utility to make it easier to share videos with my friends in our group chats, and being able to put all the complexity in something not Shortcuts feels so good.

@anisota.net's stamina system is a fresh perspective on how to interface with social media, but maybe I'm using it wrong? I think I spend too much time on each post because I've only ever depleted the stamina once. Maybe if I'm a member I can make the stamina last less for me.

Using @anisota.net slows me down so I read each post, and it exposes all the unrelated posts my feeds surface to me. I need to be more disciplined in my "likes" because even the "For You" feed is serving me too much politics.