Agentic Engineering
Region: US
Thursday 05 February 2026 15:04:09 GMT
5557
300
29
47
Music
Download
Comments
ResistForHumanity0 :
algo keep sending me this content
2026-02-26 03:13:48
2
tom :
Ai makes poor judgments frequently. I don’t trust empowering an agent because of multivariate problems and unintended consequences.
2026-02-07 14:04:17
3
The Adequate One :
Your content is always so insightful and relevant. The last time I gave Claude one of your transcripts, I told it I would love to work with you, once I have earned the right. I'll be in touch. 😄
2026-03-29 21:29:45
1
Men4Matriarchy :
Agreed. Also, in managing people, if we want them to be responsible for an outcome, we give them control over the process. The agent will pick and harmonize the processes and algorithms it is most familiar with. Step by step implementation instruction may diverge from the agents preferred (most heavily trained) processes and introduce unnecessary complications. I discuss feature sprints in depth with the agent. I then ask the agent to write detailed notes (optimized for agent) on that feature set. I then ask the agent to write up a detailed implementation plan (again, optimized for agents) broken up into phases optimized for coherence window with breaks for human testing to check the progress of the work. Part of the implementation write up is an Open Issues section, requiring me to answer vague or contradictory issues. I treat the agent as a Senior Engineer who has valuable insight about the features I am trying to accomplish. I find this process goes quite well. At times I’ve spent 60 minutes in discussion with an agent about features and implementation. Then I leave the agent to implementation, according to the plan, and it does the equivalent of 8 - 16 hors of work in the hour I’m taking for lunch. Yes, another hour of human testing and debugging is required on my return, but still a BIG win!
2026-02-22 17:09:06
2
Tazzy531 :
Huh. It seems like the same skills needed to manage humans is the same as ones managing agents. Event AI doesn’t like micromanagers
2026-02-06 03:29:23
4
tiktoker :
Not true, your architecture might not align with ai architecture of the project … sure it might work 😁
2026-02-20 14:31:25
1
Jon Capriola :
What secure what the agent executes
2026-02-21 21:10:15
1
daidaishou :
I find that giving implementation details leads to lower token churn and more reviewable code, but I can see your argument for vibe coding type stuffZ
2026-02-12 04:24:19
2
rajeshg5365 :
Doesn’t always work on hard problems. Case in point: start with a declarative goal of getting you 80% win rate at 1:1 risk to reward on your favorite stock. It’ll declare victory prematurely or fairly miserably. This does not mean that strategy does not exist.
2026-02-07 06:59:30
1
sean_salvador :
Could you give an example? Thank you!
2026-02-05 18:46:56
1
Nova3 :
You’re right about the mechanism, but the competitive advantage isn’t in the loop. It’s in what you put at the top of it.
2026-02-07 00:27:40
1
Bradford Winston Hall :
Agentic Engineering in practice incoming
2026-02-05 16:39:51
1
Agentic Engineering :
👋👋👋
2026-02-05 15:04:23
1
lazyguddu :
Each loop is a premium request. Cha-Ching!
2026-02-06 02:11:38
1
Erwin :
I wouldn't agree with you pre-Opus 4.5, but definitely now you want to get out of its way.
2026-02-05 19:16:10
1
To see more videos from user @agenticengineering, please go to the Tikwm
homepage.
