Agentic Engineering
Region: US
Monday 02 March 2026 00:26:29 GMT
6498
294
46
36
Music
Download
Comments
smile :
Yes we you describe 2 relatively similar concepts in broad sweeping generalizations then they tend to line up clearly.
2026-03-02 07:12:34
12
3x+1 :
The Post‑Symbolic Intent Layer is the emerging boundary where human intention and machine capability meet without passing through code, syntax, or symbolic intermediaries. It’s the moment when systems no longer need translation scaffolding — no APIs, no programming languages, no representational detours — because both sides operate directly in vector space, allowing artifacts to be generated from pure structure rather than instructions. This layer wasn’t explicitly designed by engineers; it’s an emergent property of scaled models, agentic behavior, and the collapse of distance between thought and implementation. It represents the shift from “writing code to build things” to “expressing intent and watching structure crystallize,” marking the first real dissolution of the symbolic‑computational divide. the things most of society and even most AI research miss are: Treating intent as a manipulable, vector-native object. Preserving heterodox, non-conforming reasoning as an explicit societal asset. Creating multi-scale cognitive scaffolding that respects individuality while interfacing with collective intelligence. Seeing physical embodiment and lived experience as critical to cognitive architecture. systems are recursive, interdependent, and sensitive to initial conditions in ways that make counterfactual stability impossible
2026-03-02 15:29:34
2
DopamineKata2 :
Are you seriously saying “using a model is a lot like using a model”?
2026-03-03 12:51:02
3
mmalpica9 :
This is a very interesting concept. I need to think it a little bit more carefully
2026-04-16 14:49:54
1
Adam❌ :
😂 💯 all my prompts are TDD.Runner(SUT)
2026-03-04 03:55:52
0
nonukiriver :
that's is an interesting interpretation
2026-03-02 00:31:46
2
Andrew :
It doesn’t look like machine leaning to me. It’s just a service that uses meta programming for control flow usually without policy enforcement for some strange reason.
2026-03-02 12:43:31
2
The crumb guy :
Recently???? Dude this was explained in depth in like the first major deepseek paper
2026-03-02 11:11:53
1
yelldha :
I could understand if someone said it in 5 years. you know it was all built on ML
2026-03-03 19:43:46
0
DogsAreTheBest :
This is why I never just let it write code. I let it write many tests of the code. Any module may have 20-40 tests that I run on the code. The only way to do this is fully separate environment inputs from business code. This then also enables the automatic finding of heisenbugs
2026-03-02 05:07:10
1
X :
The only guy that knwes what he’s talking about
2026-03-02 03:59:34
2
Carter Cole :
Lord help us
2026-03-02 18:29:19
1
Ice Station Zebra, LLC :
Intriguing. The threat envelope just got larger.
2026-03-02 02:42:28
1
Sterling :
This is so right
2026-03-02 11:27:00
1
Tom Aylott :
This is effectively what I’m working on. Yes
2026-03-02 13:46:19
1
Car mc car car :
I think overfitting is lying or cheating on coding exams
2026-03-03 02:36:22
1
3x+1 :
the things most of society and even most AI research miss are: Treating intent as a manipulable, vector-native object. Preserving heterodox, non-conforming reasoning as an explicit societal asset. Creating multi-scale cognitive scaffolding that respects individuality while interfacing with collective intelligence. Seeing physical embodiment and lived experience as critical to cognitive architecture.
2026-03-02 15:25:09
1
citharaaeternitaris :
It sounds like you're saying "agentic AI" and "fine tuning" are basically the same thing. The theoretical ramifications are interesting, but the AI pitfalls have been self-evident from the beginning.
2026-03-02 12:00:16
1
Silas 🦖 :
Why wouldn’t the asynchronous agent get referential architecture and runtime context from an existing APM/DAST. The agent learns the what and why before it is relied upon for the execution and how … i disagree … this is because most APM/DAST companies do not want to acquire agentic value props because it’s “risk” and most company’s don’t want “risk.” There can’t be model overfitting or underfitting if the foundational model is kept separate from the runtime app context (if it’s used as the agents RAG).
2026-03-02 02:48:00
1
Agentic Engineering :
🤖👋🤖👋
2026-03-02 00:26:44
0
🌊 Draper :
Gardening
2026-03-02 08:13:18
2
To see more videos from user @agenticengineering, please go to the Tikwm
homepage.
