Agentic Engineering
Region: US
Tuesday 10 March 2026 01:12:06 GMT
2220
69
14
5
Music
Download
Comments
PhloRopes :
Architecture. Information architecture, Usability / UX. SEO / AEO / GEO. Aggregated data. Planning is more important. Stop rushing. I’m guilty. You’re operating at another level. AI is like the gold rush now. Don’t follow the crowd. Step back, breathe and think.
2026-03-10 01:35:53
4
kuhlsnu :
True limitation isn’t the coding but rather content expertise. Instance pray creating a RAG without understanding how to properly make child and parent chunks and Claude it’s just likely to spit out 50 characters of data that are useless than it is to make a functioning RAG. You still have to know how to do it properly but it can code it a lot faster than you can
2026-03-12 00:53:36
1
Abbi Jain :
don't plan using prompt but use basic but written specs to make the plan. that make it so much better and ask AI to update those spec. then only you can do incremental change
2026-03-10 20:32:57
1
Kouhee :
Absolutely! For new projects can be half a day just back'n'forth to get my thoughts sorted and "edited" until the AI and I come up with an actual doable plan;)
2026-03-10 09:03:38
1
dona :
Each time I cut the lane and skipped plan mode it did awful mistakes. So I am sure planning is the key
2026-03-15 18:46:42
0
Ben :
I think plan mode is great for any major dev phase. for smaller updates, or bug fixes no need. solid documentation about design and arch constraints, properly referenced in md files, can make those incremental changes much more on target that way.
2026-03-10 01:55:00
2
Dave :
First step to ultimate optimalisation would be to implement sanitizing tools regarding data, to restrain context drift and implement fallback anchors in reasoning when interacting with the llm to scope the problem/solution in different layers of the model and/or your roadmap. Thoroughly the agent is as good as its operator and their prompt (engineering) when approaching different fields of context. Best use case for using agentic agents is a operator-first helicopterview style of approach and also configuring the agent trough interaction or documentation untill a straight line model is made for the specific task(s). AI should always be operator first. And the best approach will always be scientific reasoning. Hypothese / patch / notes / validation,confirmation / and again documentation. This way evidence always wins over assumptions. And any assumption made on memory should be placed in a specific section with also an explanation to validate
2026-03-15 21:06:31
0
Tree Luvva :
Just stick with Claude
2026-03-10 15:15:34
0
Agentic Engineering :
🤖👋🤖👋
2026-03-10 01:12:16
0
To see more videos from user @agenticengineering, please go to the Tikwm
homepage.
