March 6, 2026
Ship it or quit it
We Might All Be AI Engineers Now
Coders vs robots: learn to steer or get shown the door
TLDR: An engineer says AI assistants now do much of the coding while experienced people design and steer. Comments clash: some demand everyone learn AI or leave, others warn about flaky results and hype, with jokes about “AI slop” vs “chef’s kiss” code—this matters because jobs and quality are at stake.
An engineer says they’re designing the big ideas while AI “agents” (think tireless digital interns) crank out the code, from graph mazes to file-watchers, and even help debug faster. The claim: when guided by a clear plan, the bots write cleaner code than most humans. “That’s not prompting, that’s engineering,” the essay insists.
Cue the comment brawl. One camp cheered the easy mode future: bitwize dropped a mic with “shape up or ship out,” arguing pros must learn AI or leave. noemit added receipts: yes, AIs do dumb stuff, but they also quietly ship great code hundreds of times a day—if a good engineer is driving. ChrisMarshallNY nodded along: real value is knowing what to build and how to fix when it goes wrong.
Then came the skeptics. amelius warned that fuzzy, unpredictable systems aren’t fun when they break mid‑flight. And Bukhmanizer roasted the post itself: “sounds worse than AI slop, like ChatGPT did a line of coke,” sparking a trust debate—if the writing feels lazy, is the coding careful?
The memes wrote themselves: “AI slop vs chef’s kiss”, “copilot, not autopilot,” and “agents are coworkers, not sorcerers.” Verdict? The community is split, but very, very, loud. On balance, chaos.
Key Points
- •The author has shifted from writing most code to designing systems where AI agents and models generate significant portions of implementation.
- •A current project includes complex components (concurrent graph traversal, multi-layer hashing, AST parsing, file system watchers) with AI producing key logic under human direction.
- •Debugging is accelerated by running multiple AI agents in parallel to investigate hypotheses, multiplying the author’s problem-solving throughput.
- •Routine coding tasks such as boilerplate handlers and CLI scaffolding are largely handled by AI, increasing delivery speed.
- •The author asserts that strong foundational CS knowledge is necessary to guide AI effectively and evaluate outputs, amplifying experienced engineers rather than replacing them.