Browse Comments — Raw (as collected)

Zachary H. honestly, SO's toxicity problem was a slow self-inflicted wound. AI didn't kill the community — the community made it easy to leave

Damian Szczerba 😄 the turtle training on Splinter's own moves — that's actually a sharp way to frame it. AI grew up on SO's knowledge and now SO is the one struggling to stay relevant.

Tyrone Muhammad, BSDA exactly. SO was built for a time when finding the right answer required patience. AI removed that friction. The switch wasn't a betrayal — it was just inevitable.

Aravind S D The AI slop pushback is real and growing. If output quality drops enough, there will be a correction. But I doubt it brings people back to SO — more likely it just raises the bar for how people use AI.

more tokens for us :)

Utkarsh Maurya Utkarsh Maurya nailed it. The feedback loop matters — less public knowledge shared, weaker future models on edge cases. We might not notice for 2-3 years. Then suddenly the gaps show.

George Coller M.S. I'm glad you've found where your abilities complement the AI system. Many are still struggling with this!

kinda like how CAD tools shifted architects from drafting every line to really exploring forms, AI lets researchers skip routine stuff and dive deeper into new ideas.

No comment.....

Stack overflow is no longer relevant. Remember the time humans transitioned from machine language to programming? Now we are transitioning from programming to ai tools. Programming languages are irrelevant now. They will slowly become obsolete. Now skills are measured by how learnt you are in claude, cursor, chatgpt etc. If you are still coding the traditional way or even debugging the traditional way, you are going backwards. AI should code and debug for you henceforth.

Sometimes the logs in third-party libraries are so cryptic that I don't think we need to worry about losing our debugging skills anytime soon! And that Stack Overflow nostalgia... Oh:)

Very good reflection. During my first experiences using AI for software development, I quickly realized that although I was producing code faster, I was not really learning. That’s when I decided that for every new project using a framework I don’t know, I would build a complete progressive workflow — from beginner to professional. The project evolves from a beginner version to a professional one by following clear stages that I test at every level. I do not focus only on the code itself, but on understanding the logic behind how the entire project works. During these progressive checkpoints, I continuously update the workflow with best practices, solutions to bugs encountered, and lessons learned along the way. At the end of the project, I create a second, professional-grade workflow for future projects. The first serves as educational material, while the second becomes a production-ready process for similar projects. If you have other approaches that you use, please share them with me — I’m always interested in learning different ways of working and improving my process.

“STACK”

Wowwwwww

Thanks Qi Han Wong, very interesting!This maps very directly to legal AI too. Language is not jurisdiction. A Spanish prompt may require Argentine, Spanish, Mexican, or US law. An English contract may still be governed by Argentine law. If the model silently treats language as a proxy for geography or governing law, it may understand the risk correctly but route the answer through the wrong institutional pathway. For legal and compliance AI, explicit jurisdiction anchoring is not a detail. It is a safety layer: governing law, forum, user location, institutional authority, and role of the user all matter.

Spent time recently auditing how many tools sit in a typical AI workflow and the number was embarrassing. The switching cost adds up in ways that are hard to measure but easy to feel. A unified workspace solves a real problem, not just a convenience one, Paul Storm

Stack

Luís, I’ve seen this "human anatomy" analogy all over LinkedIn lately. It’s a clean framework, but it dangerously oversimplifies the reality of high-stakes environments. Bridging bedside medicine and AI architecture, I see a major flaw: this body is missing an immune system. In a hospital, "agents with hands" operating without strict, fail-closed deterministic gates is a recipe for fatal never-events. The clean logic of enterprise AI always shatters against the chaotic, noisy reality of clinical workflows. We need less hype about "brains" and more focus on "immune responses." I'd value your thoughts: 1) Technically, how do we architect this deterministic "immune system" into the MCP layer to intercept hallucinated agent actions? 2) Clinically, how do we stop this "brain + hands" hype from pushing leadership to deploy autonomous systems before EHR data is actually clean?

STACK

black humor