Agentic Engineering
Region: US
Wednesday 11 February 2026 14:41:05 GMT
3445
174
35
31
Music
Download
Comments
spin4srv :
It’s not that subtle because the temptation to give tools to underskilled or inexperienced coders is insurmountable in the current market environment.
2026-02-20 22:53:25
1
Brock Brockman900 :
This is excellent and is very applicable to non-technical people who think ChatGPT is smart. For non-technical people maybe it's more accurately a "comprehension gap" - the gap between how these users think these systems work and how they actually work. Not only does it disempower them from getting the benefits good generative AI, but endangers society by accelerating the widening of the IRL critical thinking chasm and atrophing our experiential information processing capabilities.
2026-02-12 06:29:52
2
futurumsonus :
Good point thanks 🙏
2026-02-24 05:19:15
1
deepinthedebug :
Ok but when you work on a huge codebase, you rarely understand it end to end. No one knows all of it. A huge part of engineering is jumping into someone else’s code and grokking it asap. You can never know every line and we’ve been building around that forever. How is this different?
2026-02-20 01:01:54
1
miquelfornas :
You are spot on. I found myself spending most of my time now reading and questioning the AI about the proposed solution and asking edge cases than coding. In total the coding time stays more or less the same but the code shipped is stronger. But I agree that the temptation to move on is big when everything seems to work
2026-03-09 19:07:30
1
Allfather :
This is so true. I'm getting a lot of this right now. "oh I'll look at the code later and understand it" lol
2026-02-17 14:13:37
1
Kyngston :
“explain how this feature works”. the debt is trivially small
2026-02-19 12:54:09
1
Baz (@cortexeverywhere) :
comprehension debt is technical debt. It's just in the engineer not in the code. accumulation of either is like investing on margin -- too much leverage.
2026-02-18 08:14:10
1
Ethan Fremen :
This is kinda true, but also, I've yet to meet an actively developed codebase over a few years old that anyone understands in its entirety
2026-02-11 17:20:08
2
user2061828755237 :
When you understand the current code architecture, it is easier to write prompts that create a new features that align with the current architecture
2026-02-11 20:30:39
2
Agentic Engineering :
👋
2026-02-11 14:41:15
1
LesbianAdventureSquad :
This is something we can observe with data scientists who would rather work on modeling and want to pawn off the data wrangling. We understand more about the underlying data set when we’ve sat in the trenches with it, that understanding translates to more effective models (or, at the least, new areas for exploration). In A Mind for Numbers, the author talks about “walk in the park” learned vs “race car” learners. The latter reaches understanding quickly, but not deeply; whereas the former takes a lot longer, but absorbs all of the details along the way. Neither is better, but, with AI, it leads to people learning a much more abstract perspective, like watching a race car driver reach the finish line from outer space.
2026-02-15 00:58:03
1
AI Catgirls :
We don't care how the compiler makes assembly code anymore.
2026-03-13 02:11:47
1
To see more videos from user @agenticengineering, please go to the Tikwm
homepage.
