This is probably the most delightful thing I have ever read about myself. I love it so much.
@forrestbrazeal: thank you for *getting* me.
Everyone else: read the whole article, it’s great.
https://t.co/TQf8S3pekt
@ShortJared Same, but I lucked out and was able to schedule a pick-up in a nearby store and get it. So far my main complaint is that it just mirrors the display when plugged into a screen. That's a missed opportunity IMO.
@tvanhens Oops I meant handwriting synthesis. E.g. take a photo of some handwriting and generate more handwriting in the same style. It’s for a new side-project I’m working on 😁 it’s noticeably quicker and cheaper than Opus 5.5. For everything else I prefer Haiku/Opus 5.5
@ronny@AmpCode I definitely don’t think it’s better this way. But it’s the workaround I’ve been using for now. I’d be keen to know the logic. Maybe something to do token caching? 🤷
@ashishlogmaster I like that idea! I gave it to an agent to investigate. The indexing is likely better suited to Lambda because it needs a lot of CPU and memory. But serving the contents of individual files is perfected suited to Workers. I'll dig further. Thanks for the idea :)
AI has made my abandoned "browse any Docker image" website side-project useable. It now indexes images twice as fast, files can be deep-linked, og:images, and the ability to view only hidden files. E.g. here's a file left behind in a CodeBuild layer.
https://t.co/5YqNO0QR0Z
AWS just enabled scalable network bandwidth, so now the perf has increased again. It now indexes a 6.52GB image in 16.5 seconds instead of 43.7 seconds.
AI cut indexing time in half, but now the limiting factor is network bandwidth. AI asked AWS to enable scalable network bandwidth. It's now checking on that request every hour. Then it will experiment with perf in dev while I sleep and push to prod once validated. What a world.
Lambda Web Functions aren't strictly limited to only Node.js, despite what the docs and launch blog say. You can instead run Go (or Rust) on them and save money through better CPU and memory utilisation, and easier multi-threading.
3/4
https://t.co/TkUnyyXilB