bentcorner

Kind of makes me wonder how knowledge transfer is going to work in the future.

How many times have you found a random forum post solving your problem? How many countless people has that post likely helped?

Now every single person facing the same problem will regenerate from scratch the entire thought process and result. What a waste.

There needs to be some way to close the feedback loop from the work that LLMs do.

How do you feel about signing in through Muse? The first time I asked it something that was only available in Maps browser it popped the browser and asked me to login to google and it just felt wrong. How secure are their VMs?

This kind of approach works so well.

I was trying to upload HDR photos to Instagram and couldn't figure out how to make it preserve the HDR-ness of my photos, and was going back and forth with an agent - step by step we'd slowly iterate on a photo, I'd upload, then I'd download the jpg and give it to the agent and it'd think, change something in the format, and we'd try again.

I got impatient and just said fuck it and gave it chrome-devtools and a blank instagram account and a test photo, and it figured the whole thing out in about a half hour. I built a little browser-based tool from what it discovered, and I use it to run my HDR photos through to format them correctly for instagram and it works a treat.

My work provides me a Mac laptop plus remote Windows VMs and I've completely stopped using my Mac for agent builds/tests - I give them ssh access to WSL instances and they can go hog wild on the VMs.

It's all a bit adhoc - agents still need to pull (or I need to keep the main enlistment up to date on these machines) - I'm thinking of using containers or something else like github actions keep things cleaner, but right now it's "good enough".

Regardless I'm pretty happy with how it's working so far. Previously I'd be letting agents do work in the background and they'd kick off builds in the middle of a meeting or something and I'd start dropping video streams until I killed everything.

I use an agent to author my nix config - personally I don't have the determination to have picked Nix up entirely on my own so I find it to be a godsend. I can code review config changes and keep it all in git - also makes it easy to pick up the entire config and deploy it on a different machine.

And dropbox is just rsync and some UI

At one time I heeded advice not to rely on 2FA SMS because you know it's not secure, so I kept everything in a 2FA app.

Well through an unfortunate sequence of button clicks I wiped my 2FA state and had a hell of a time getting into the apps that I removed my SMS from.

In theory yes SMS is an attack vector but personally the safety and convenience of SMS is more valuable than the odds that someone impersonates me at my mobile provider.

(Also I stopped using that 2FA app and just use a keepass db stored in onedrive).

This is a habit tech people fall into. "Amazon deleted my book" -> well just strip the DRM off it. "My ISP monitors me" -> well just use a VPN instead. "Ads make it hard to use the internet" -> well just use an ad blocker.

Ironically if more tech people just rawdogged the internet I think we would have more progress.

Spotify works mostly fine for me but there's weird rough edges that I don't always see in "simpler" apps.

Stuff like transitioning between desktop and mobile - it tries to work and most of the time it does, but sometimes I need to manually make my desktop play on my desktop. Not a big deal but sometimes I'd be happier if this feature didn't exist at all.

On my desktop, Spotify can be finicky when I connect/disconnect my bluetooth headphones. Sometimes it refuses to play after I connect my headphones, and I need to restart the app. I think it's trying to do something magically for me after it detects the headphones and gets into a wedged state, but a simpler "play to active speaker, whatever it is" audio app wouldn't have this problem.

This happens to me too and I enable 'Live Captions' (not meeting transcriptions) so that I always have a scroll back in case I miss something.

It's built into the meeting software I use and probably in most mainstream ones.

I tried OS transcription but those don't split out who's talking (since it's a single audio stream) so much less useful.

Anyways all that to say that at least in online meetings this problem is solved and addressing this problem in a stack that doesn't have meeting context is going to result in a worse experience.