Claude is only sycophantic until the bill comes.
#GOFH
/E@A:SIU¼DRJNFCKTZLWHYPQOBGMXV£
| BLOG | https://clayto.com |
| GITHUB | https://github.com/mwcz |
Michael Clayton
@mwcz
- 21 Followers
- 55 Following
- 147 Posts
Programmer at Red Hat.
#GOFH
/E@A:SIU¼DRJNFCKTZLWHYPQOBGMXV£
#GOFH
/E@A:SIU¼DRJNFCKTZLWHYPQOBGMXV£
@fasterthanlime @words_number Ah, alas for the friendship we could have had!
@words_number @fasterthanlime what he said 😅
I was thinking about Turing's audio debugger and it made me wonder if anyone's applied that in a modern debugger. Either audio pulses or, more practical, a small sparkgraph-style indicator showing breakpoints hits over time. Disabled breakpoint "hits" would be neat to visualize, and would give them some utility.
In video calls, nobody will know if you turn them into chipmunks by raising audio output by an octave or two. (using EasyEffects, etc)
@danielsreichenbach I've used heaptrack, or valgrind's massif. Test binaries are inside target/, so you can run those tools against the binaries. It's not always easy to find which binary, since there's one for each crate, and one for each integration test module, iirc. Hope that helps.
BEEP BOOP. Trepidation about Claude Code melts away after telling it "make robot noises when you make a mistake". BZZT!
@bcantrill @ahl Looking forward to it! I'm especially curious to hear ways to stay rigorous when reviewing LLM output from others. Whether it's PRs, job applications, documentation... the sheer volume of these things is increasing rapidly and staying rigorous is getting more challenging.
@mwichary If I had to bet, my prediction is that garish ads will be added to the free tiers, similar to YouTube & almost every free service on the Web today. But free and paid tiers will have something more subtle. Not "ads" but biases towards certain products, services, etc. Asking chatbox XYZ for shoe recommendations will reliably result in Nike, that kind of thing.