estop
Region: US
Friday 07 August 2026 22:30:44 GMT
16131
323
28
19
Music
Download
Comments
The AI Syndicate :
You shouldn't be babysitting your AI. You should be setting up human in the loop gates at the crossroads of "Expensive to continue a mistake", and if continued "Hard to catch later." Only when those two things are true should you be actually checking the work. A good example is before you publish AI work. A mistake is costly and could cost your reputation or lack of sales, and once posted and you continue moving along with your work, that mistake isn't something you'll easily find until it bites you. Stop babysitting every step. If AI makes a mistake, have a checkpoint where you look and fix it. Its ok to let some mistakes get through to your checkpoint. Youll save more time in 15 minites of review\editing at the gate you create than making sure each step is error free.
2026-08-08 14:33:19
1
Faithless? :
--dangerously-skip-permissions is peak 🙏
2026-08-08 08:14:07
8
Voxaris :
I use all of the different AI’s pretty much daily, but I don’t think that anyone cares more about their customer base or has a loyal customer base like Anthropic. I feel like they really put a lot of effort their base of consumers.
2026-08-07 22:36:05
12
merenotnoone :
89% ? that's not good wth
2026-08-08 01:17:14
1
rad :
How clearly?? If it’s really as clear as they make it out to be 89% isn’t that much of a flex
2026-08-08 01:49:39
1
Joe Average :
Imagine being swapped from fable to lower model because the stupid thing though you building something dangerous for a simple app health... It's getting extremely stupid to pay for something I can run locally
2026-08-08 01:40:59
3
Christopher^ 🌼 :
Im not really a fan of AI. But what I do if I need to use Claude code is I only use it in a dev virtual machine. It has permission to run dangerously and it can make its own commits without my input. What I do from there is I do git fetches / cherry picking. I’ll audit the code and apply my own edits. And if I’m happy, I’ll commit it as my own. I would never give Claude code or really any AI direct access to my work station.
2026-08-08 09:49:39
1
D4m4s74 :
turn Claude off
2026-08-08 08:14:13
1
🆁🅴🆉🆇🆃 :
doesn't really matter as long as it doesn't stop the model from performing actions, explains why it stopped it and gives you a choice. if it doesn't break the agent loop then model will just find a workaround solution for the same task that is possibly worse and collects garbage data on the machine
2026-08-08 07:08:20
1
estop :
claude is the 🐐
2026-08-07 22:33:08
2
Sushii69 :
i dont have time to sit for every confirm
2026-08-08 08:42:16
0
AllanatheGreat :
Thanks Schlep!!!
2026-08-08 05:17:33
1
Johny C :
Bypass permissions and go to sleep
2026-08-07 23:10:03
1
Rhett.Richter :
🐐🐐
2026-08-07 23:49:23
1
To see more videos from user @estop845, please go to the Tikwm
homepage.
