The Hype Kevin
Region: CA
Sunday 18 October 2026 17:00:00 GMT
38549
1558
33
195
Music
Download
Comments
LucFlow :
Bro thinks I’m checking 40
2026-10-18 22:44:16
50
Dru Steeby VO :
This approach cannot work for mission critical software. Claude code is by definition non-deterministic. If pushing our code straight to production caused a manufacturing plant to go offline for even an hour while the agents debug it can cost millions.
2026-10-18 21:44:08
33
Marty Doyle :
Write your tests first
2026-10-19 04:02:18
1
AugustusFinknottle 🦎 :
"Comments that are only about style get dropped." "Style" is what makes code readable, modifiable, and debuggable. This is going to bite them on the ass down the road.
2026-10-18 22:39:46
6
.. :
that's a lot of tokens
2026-10-18 21:40:49
11
Elideus :
Hell yes do not let the AI write the CI lol :)
2026-10-19 01:37:43
0
aldov500 :
User is prod are checking the code bro don't worry
2026-10-18 21:19:57
40
joshsmith9717 :
And has anyone noticed a material difference in claude code in the last 3 months? The last 6? 12? Maybe one or two features? This is madness
2026-10-19 00:32:37
1
Tinsu :
I’m curious whether PersoWrite can make a draft feel less repetitive without removing details that matter
2026-10-18 19:05:53
1
Fox :
you think we need to trust the code review skill from claude or we must have our own skill aka custom agent. also the other model for reviewing the code should be a higher model or it's not a problem a lower model e.g. from Opus to Sonnet
2026-10-18 17:10:47
0
Elideus :
Repeat three and reconsider approach is a good tip
2026-10-19 01:36:56
1
kennovation :
It’s okay to have AI write ALL your code just do it in small pieces like who tf is making 600 line PRs bro
2026-10-18 23:25:45
3
executeorder66 :
this is a great story
2026-10-18 19:53:06
0
trial_and_errored :
I’ve been building an at home version of this since last November with open source models, and customizing a harness to be security forward. I didn’t know it was the same model that Anthropic is working with. However- after watching your video I’ve got some work to do: the “who decides if a person” rounds by finding type was missing. I’ve been manually reviewing sensitive paths, and planning for deep hands on texting before product releases. Seriously, thank you for this breakdown. I’ve been too heads down to research outside of this scaffolding!
2026-10-18 22:05:46
1
mati69pl :
i feel like this couldve just been done with gerrit but they wanted the "new git" headline
2026-10-19 00:09:08
0
Markizano Draconus :
Agents reviewing agents is a disaster designed to max LOC counts as vanity metrics and call it productivity...
2026-10-18 23:22:00
0
cc :
skynet is near
2026-10-19 02:06:43
1
To see more videos from user @thehypekevin, please go to the Tikwm
homepage.
