Brock Hart | Relational AI
Region: CA
Wednesday 21 October 2026 15:29:07 GMT
39691
1034
66
90
Music
Download
Comments
BenSmidt :
Just have AI read the 3 books so you can trust the output a
2026-10-29 00:26:00
0
Adonna :
AI talks too much tbh. It's way too wordy. Too many blank lines. It's not concise enough to be efficient.
2026-10-22 00:49:18
130
Beanz999 :
Just use AI to review AI and tell it to make no mistakes
2026-10-22 01:28:48
9
WeWalk :
never review
2026-10-28 19:02:55
0
Emo Millennial :
AND you still absolutely need humans in the loop. nothing AI produces is ready on v1.
2026-10-23 02:36:06
9
Ροδάμανθος Ανδρουλάκης :
you need to hire more people to review
2026-10-28 17:26:49
3
djhead73 :
My car can go at 150 MPH, I'm holding it back.
2026-10-28 19:33:56
2
Moneymaxxer :
Required human judgment is already on a exponential downward trend. 3 years ago I had to double check a simple class refactor. Today I can let it own a feature. I am confident it will take over all development and digital operations within next 3 years and hopefully help me retire soon after 👍🏼
2026-10-21 20:14:07
8
intergalactic steamliner :
It’s wrong more than it’s right, and it takes me longer to edit its work than if I had just written it myself to begin with.
2026-10-23 20:28:50
1
Vintorzaleris :
what? the thing studied itself and and reported back that it's better then everyone else? wow!
2026-10-22 14:07:04
7
Sentinel :
faster, quantity, and capable is not the same as quality, proper judgement and business ready. AI still spits a lot of errors. Plus, I wouldn't trust my data or clients data to it, anytime soon. PHP sites hack? what?
2026-10-22 12:53:27
3
Nat Cat Eye :
Wow 😳😳😳 that's important. This Is how we slow down AI
2026-10-21 20:28:45
5
user4423472714366 :
yeah, and a neverending review cycle of code nobody understands end to end, and thus cannot succinctly explain to a reviewer is exhausting. we're going to see some brutal burnout on these cycles if we dont figure out how to bridge that gap
2026-10-23 02:46:31
3
Chris :
Reading three books to verify AI content is irrelevant if AI was involved in any part of the production of the books. We are setting ourselves up to exist in a world where verification by a human agent is no longer even possible, putting aside the question of whether it will be relevant
2026-10-22 08:19:22
3
mrsanchez303 :
The technology is incredibly powerful, but you still need to know what you’re asking for. You need the knowledge and experience to understand how something works, what you want to create, and how you envision it functioning. Without a foundation built through real-world experience, education, reading, and years of learning, you’re only scratching the surface of what AI can do. AI is a tool, but your knowledge, experience, and vision are what unlock its true potential.
2026-10-22 00:54:32
2
Scott (Jennifer’s Version) :
Feeling this daily. But it’s a new type of slop. We can generate much more now, but it doesn’t make it good when it’s based on what we know. So we’re creating automated curation, but that’s also based on precedent. Can’t judge new if we only make our own versions of what already exists. We’ll soon realize we way overindexed on sameness, and that’s when the real innovation comes for regular everyday humans with AI.
2026-10-22 01:46:16
2
Birb. 🦜 :
Yeah because when an AI gives a “summary” it’s like a 6 paragraph essay. They cannot shut up 😂
2026-10-22 06:29:55
1
Ashli :
Yeah my buddy says the only way he can move faster is if he can read faster
2026-10-22 01:30:04
3
bijonkijank :
At the end of the day, nothing gets made without human labor, no matter how much of the process is automated
2026-10-22 02:58:28
1
amackintosh 🇨🇦 :
Don’t forget this is only temporary. It won’t be long before we don’t even review it anymore
2026-10-21 23:51:08
1
tooba_the_hutt :
Number of agents does not mean they are producing more than humans can review. It means they’re getting projects done faster. Very different.
2026-10-23 21:08:44
0
Mustapha :
I had exactly this issue today, a colleague started pumping out PRs with hundreds lines of code, and got mad that they took too long to review!
2026-10-28 17:50:21
0
TheEternalBruh :
If it’s making too much work, have it spend some % of its tokens on making explainer diagrams, slides, videos, interactive walkthroughs. Make it make reviewing easier
2026-10-27 21:21:07
0
Imagile :
We should design work to have agents prove their work to humans; having it code a feature is one thing, but presenting evidence and test harnesses to show it, optimize PRs for human review, will make it even faster - if human attention is the bottleneck.
2026-10-24 17:26:31
0
J.R. :
LLMs don't need to be perfect, I think people are not bringing that point up. What process they need that matters is fast recovery of an issue occurs.
2026-10-27 11:52:28
0
To see more videos from user @overlappingbrock, please go to the Tikwm
homepage.
