Dan Dean
CI wasn’t passing so someone’s Cursor agent just went ahead and changed the CI workflow 🫠
Mar 13 at 4:42amWeb
Show threadkfMar 13
@dandean this toot isn't real it can't hurt me
Show threadDan DeanMar 13
@kf I’m excited for these technologies to be applied to actual fields of engineering.
Show threadDan DeanMar 13
@kf the bridge keeps collapsing in these test scenarios. Better change the tests!
Show threadkfMar 13
@dandean no tests only vibes
Show threadLeigh CaplanMar 13
@kf @dandean clearly AI has watched too much MacGuyver
Show threadkfMar 13
@texel @dandean worse: it hasn't watched macgyver, only read the captions
Show threadEvan StonerMar 13
@kf @texel @dandean kf is like “you take Richard Dean Anderson’s name out your god damn mouth”
Show threadkfMar 13

@estoner @texel @dandean did I ever tell you I used to have an RDA fan tumblr when I was a teenager

kf lore

Show threadEvan StonerMar 13

@kf @texel @dandean you did not, but I have exactly zero surprise

It occurs to me that I was a big MacGyver fan at roughly the same age as you would have been a Stargate fan. RDA is for the children, of every generation

Show threadPseudo NymMar 13

@dandean

Test fails.

Delete test.

No more failure.

Show threadTad FisherMar 13

@dandean why I advocate for "agentic polyrepos", a concept I just pulled out of my ass like karpathy

code goes in one repo, tests go in another, ci yet another. your agent only gets write access to one at a time.

surely this will 10x productivity

Show threadEvan StonerMar 13
@tad @dandean “ass like karpathy” is now in my head to the tune of “moves like jagger”
Show threadTad FisherMar 13
@estoner @dandean sorry I meant ass-like karpathy
Show threadTheZwick32Mar 14
@dandean
That defeats the entire purpose of CI.