Andrea Matwyshyn

@andreamm
1.6K Followers
397 Following
462 Posts
Professor, law and engineering, Penn State; Founding Director, PILOT lab & @manglonalab
My labshttps://www.pilotlab.org and https://manglonalab.org
Websitehttps://www.andreamm.com
Birdsite@amatwyshyn
Wikipediahttps://en.wikipedia.org/wiki/Andrea_M._Matwyshyn
Sometimes squirrels jump out and find you when you least expect it... sometimes even in the middle of an ML talk.
Still waiting for a service capable of generating infographics that correctly spell words *as I have typed them* in the instruction. We are not there yet.
On the Internet, everyone knows you're a dog.

My adventures in vibe coding continue. This time I was attempting to build a simple text to voice app with manual pronunciation correction, extra pausing, and downloadable .WAV file generation. I have ended up in the same place as I did six month ago with a cite generator project: debugging by hand and with only one potentially usable app, despite trying multiple platforms.

Here's the humor winner for the best Console of the night. ....Same, Console, same.

Chef's kiss.
Six finger problem, meet the five branch problem.

Also, holy fake endorsements, Batman... As expected, accuracy of information was problematic.

Aspirational vibes of AI functionality != fitness for purpose

""Use our AI to redesign your website in the style of your favorite site!"
<Hmm, ok let's see what it can do with my personal site>
"Meet our AI CEO!"
<ok, whatever...>
"Check out your new design!"
<drumroll>
"The page encountered a (sic) error..."
<not good... but ok fix>
....um.... THAT IS NOT MY NAME...

Proof, pudding, self-described one job, AI hype, etc. etc...

This does not instill confidence in the future of vibe coding.