@thehypekevin: You can only keep track of a few agents at once. DeepSeek's fix is to take the lead's hands off. So does it work on real bugs? More breakdowns on YouTube, link in bio

The Hype Kevin
The Hype Kevin
Open In TikTok:
Region: CA
Wednesday 04 November 2026 17:00:00 GMT
32175
951
23
253

Music

Download

Comments

hwfcup
nazar :
I was running this setup with opus 4.6 back in the day and it was very slow, and quite expensive. Just manage the sessions, don’t let them talk to each other
2026-11-06 12:38:21
1
hearseefeel
HearSeeFeel :
I have a Claude session running in yolo mode with herdr and it approves anything it wants In Other sessions
2026-11-09 06:04:24
0
boecht
boecht :
That's revolutionary! If you now combine it with status, let's call them backlog, in progress, review, and done, this would be a radical setup for late 2024!
2026-11-04 18:02:30
14
ssbotz
Shravan Suryanarayan :
I have a database in my project and not really comfortable running multiple agents as they may override the data or tables. how to handle it
2026-11-04 21:08:50
2
bob.smith1039
Bob Smith :
This is near enough my setup I have but I focused it around GitHub Projects so there is an intake, grouping and ordering of issues and then commitment of tasks all sent to a pool of agents with drift checks in place to avoid things going wild and keeping it aligned on the projects architecture and goals.
2026-11-05 04:38:53
0
landon_bee
Landon B :
Bro those some amateur numbers haha. I have 24 main agents 😏
2026-11-08 05:45:49
0
little_ms_abi
another feminist 🏳️‍⚧️ :
I set this up with custom agents. one orchestrator, one analyst, N developers, one tester. I signed different models for each agent. opus for orch/analyst, sonnet for Devs, haiku for tester. a solid phased spec is key to implementing this way. auto confirmation and merging for sub agents with orchestrator validating on task completion.
2026-11-06 07:40:29
2
yonatannaor771718
yonatannaor771718 :
I’ve built an agent orchestration where each project has a manager session that can start sub sessions (not sub agents) they can talk between them and the manager can answer the questions of the stuck sessions
2026-11-06 08:21:35
1
nusuntnormal
Dragos Catalin :
I have about at least 20 individual chat sessions each with a maximum of 12 sub agents
2026-11-06 03:54:04
0
_orange_triangle
🟪 Orange Triangle :
Just open grok in its own terminal and tell it to say yes to anything waiting .. optimize and keep any dialogues like the ones that open from opening in future then start as many terminal windows as you want
2026-11-06 03:41:21
0
thatmilwaukeeguy
ThatMilwaukeeGuy :
this is what the new persistent agents do, like the dot feature on openai. it is seemingly life-changing, but the drawback is that it actually can slow down progress because it is orchestrating and you don't have control over how fast it checks in on the various tasks that go into long running mode. on complex efforts, I'm finding that my ease of use has increased significantly but my output has actually decreased because I was faster working directly with the sub agents myself. just a hell of a lot busier
2026-11-04 23:10:50
1
aaronpereza707
Aaron Perez856 :
For me it’s always the way around. I give the main agent the possibility to use subagents and it always does. Once I spent my 5 hour limit like in 20 minutes haha So now, for every big task, I need to specify no more than 5 subagents. No subagent can have the same model and thinking effort as you. You are, let’s say fable 5.1 and the most a subagent can be is opus 5.5 high. You don’t do the work you orchestrate and receive updates, spawn auditors to check each agent’s work and so on.
2026-11-05 15:01:51
0
jaipears2066
jaipears2066 :
Wow. Happy someone actually explained it well. Excellent content.
2026-11-04 23:51:45
1
classicalman333
James Barris :
Too long, make it shorter and succint
2026-11-05 01:11:29
0
1_tec
Thomas :
me 2
2026-11-05 04:48:10
0
To see more videos from user @thehypekevin, please go to the Tikwm homepage.

Other Videos


About