How People Are Actually Using Agentic AI

That exhausted look when your coworker won't shut up about AI. The quotation marks around "programmers" say it all - real devs are busy fixing merge conflicts while the ChatGPT enthusiasts are planning how AI will write their next project. Meanwhile, the rest of us just want to finish our coffee before it gets cold.

Ah, Amazon Kiro's "revolutionary" spec-driven development in a nutshell! A fancy Venn diagram showing "the flow of vibe coding" and "the clarity of specs" with a ghost in the middle representing the haunting emptiness of this approach. But peek under the hood and what's the grand innovation? Just concatenating the spec to the prompt. That's it. That's the revolutionary breakthrough. The shocked cat at the bottom perfectly captures every engineer's reaction when they realize they've sat through a 2-hour meeting about a "groundbreaking methodology" that's just prompt = spec.toString() + prompt . Corporate development in its natural habitat – repackaging the obvious as innovation since the dawn of computing.

Reddit programming subs in a nutshell: GitHub Copilot adding over a million lines of code while removing just 332. Then there's the "vibe coders" adding 153K lines but deleting 9K. This is the digital equivalent of that coworker who writes 500 lines to do what could be done in 10. Sure, the git stats look impressive, but someone's gonna have to maintain that monstrosity after they move on to their next "10x developer" gig. The real heroes are the ones who commit -5000 lines that make everything run twice as fast. But they don't get Reddit karma, do they?

Anthropic really out here offering $570k/year for a Software Engineer role that "may not exist in 12 months" because they know Claude is about to automate everyone out of a job. The irony is chef's kiss—they're basically saying "hey come work on the AI that'll replace you, here's half a mil for your trouble." That disclaimer at the bottom hits different when you realize they're not worried about funding or pivots... they're worried their own product will make the position obsolete. Imagine putting that on a job posting. "Join our team to build the thing that makes your team unnecessary!" At least they're honest about it, I guess? The real kicker: someone's gonna take that offer, bank the cash for a year, then use Claude to build their startup while unemployed. Circle of life.

The iceberg of AI terror is real, folks! On the surface, it's just "AI" - those fancy chatbots everyone's talking about. Dive a bit deeper and you hit "Machine Learning" where your code starts making decisions without you explicitly telling it how. But the true horror? That murky "Deep Learning" zone where neural networks do their black magic. And what's holding this entire technological monstrosity together? Some poor developer's spaghetti Python code and linear algebra that they barely remember from college. The whole industry is basically running on StackOverflow answers and caffeine. Next time someone says they "work in AI," remember they're just the tip of an iceberg floating on a sea of mathematical duct tape and prayer.

Brutal honesty right here. Everyone's building "AI-powered apps" but let's be real—most of them are just fancy UI layers slapping a markup on OpenAI API calls. You're not doing machine learning, you're not training models, you're literally just buying tokens wholesale and reselling them retail with some prompt engineering sprinkled on top. It's like calling yourself a chef because you microwave Hot Pockets and put them on a nice plate. The term "wrapper" at least had some dignity to it, but "Token Resellers" cuts straight to the bone—you're basically a middleman in the AI supply chain. No shade though, margins are margins, and someone's gotta make those API calls look pretty.