georgemandis

5 ms·

1.
by georgemandis 15d ago
Starlight | Senior Full-Stack Engineer | REMOTE (US), NYC preferred I’m the Head of Engineering at Starlight, and we’re hiring a senior full-stack engineer. We build software that helps financial institutions connect people with government
2.
by georgemandis 4mo ago
I'm realizing `encodeURIComponent` is actually part of the ECMA spec! I thought it was something provided by the browser like `window` or `document`. I withdraw my "the language handled it fine" comment, haha. Before I'd
3.
by georgemandis 4mo ago
>I do not agree that slice() should operate on extended grapheme clusters. Don’t lump the grapheme cluster/scalar value split in with the sins of UTF-16 and its unreliable code point/code unit split. Yeah, I think that's f
4.
by georgemandis 4mo ago
You are reminding me we also circled an issue at one point where a backend system in Python needed to agree on the same character count length of a piece of content was the client (JavaScript). Another place Intl.Segmenter would've hel
5.
by georgemandis 4mo ago
The language handled it fine. It will generally just show replacement characters (�) for combos that don't map to anything. It was really `encodeURIComponent` that didn't handle it gracefully. If you just type this into the consol
6.
by georgemandis 4mo ago
My recollection (that I didn't add to the story): I don't think Intl.Segmenter had great browser support then (2022). Even if it had it still wasn't a quick/obvious fix for our problem with where it was occurring in our
7.
by georgemandis 4mo ago
Just noticed this is getting some traffic! It's a little buried in the post, but I made an interactive tool for exploring surrogate pairs as part of this: - https://george.mand.is/invalid-surrogate-pairs/ I though
8.
by georgemandis 1y ago
Definitely in the same spirit! Clearly the next thing we need to test is removing all the vowels from words, or something like that :)
9.
by georgemandis 1y ago
I had this same thought and won't pretend my fear was rational, haha. One thing that I thought was fairly clear in my write-up but feels a little lost in the comments: I didn't just try this with whisper. I tried it with their new
10.
by georgemandis 1y ago
I kind of want to take a more proper poke at this but focus more one summarization accuracy over word-for-word accuracy, though I see the value in both. I'm actually curious, if I run transcriptions back-to-back-to-back on the exact sa
11.
by georgemandis 1y ago
I watched your talk. There are so many more interesting ideas in there that resonated with me that the summary (unsurprisingly) skipped over. I'm glad I watched it! LLMs as the operating system, the way you interface with vibe-coding (
12.
by georgemandis 1y ago
Hahaha. Okay, okay... I will watch it now ;) (Thanks for your good sense of humor)
13.
by georgemandis 1y ago
Interesting! At $0.02 to $0.04 an hour I don't suspect you've been hunting for optimizations, but I wonder if this "speed up the audio" trick would save you even more. > We do this internally with our tool that automa
14.
by georgemandis 1y ago
Yeah, I'd like to do a more formal analysis of the outputs if I can carve out the time. I don't think a simple diff is the way to go, at least for what I'm interested in. What I care about more is the overall accuracy of the
15.
by georgemandis 1y ago
Should be fixed now. Thank you!
16.
by georgemandis 1y ago
For what it's worth, I completely agree with you, for all the reasons you're saying. With talks in particular I think it's seldom about the raw content and ideas presented and more about the ancillary ideas they provoke and i
17.
by georgemandis 1y ago
Oooh fun! I had a feeling there was more ffmpeg wizardry I could be leaning into here. I'll have to try this later—thanks for the idea!
18.
OpenAI charges by the minute, so speed up your audio(george.mand.is)
740 points by georgemandis 1y ago|228 comments
19.
by georgemandis 1y ago
I was trying to summarize a 40-minute talk with OpenAI’s transcription API, but it was too long. So I sped it up with ffmpeg to fit within the 25-minute cap. It worked quite well (Up to 3x speeds) and was cheaper and faster, so I wrote abou
20.
Ask a computer A toy powered by GPT-3 and reckless abandon(george.mand.is)
1 points by georgemandis 4y ago|0 comments
21.
New Elmo fire memes with DALL-E 2(george.mand.is)
2 points by georgemandis 4y ago|0 comments
22.
npm install turboencabulator(george.mand.is)
1 points by georgemandis 5y ago|1 comments
23.
Facebook Recruiter Correspondence(george.mand.is)
156 points by georgemandis 5y ago|146 comments
24.
by georgemandis 17y ago
Why not just release it as an open-source project?