AI or You? Who is the one who can't get it done? (medium.com)

AI isn’t simply “good” or “bad” at tasks — this piece argues the gap is mostly driven by the user. The author reframes common polarizing stories (AI hype vs. doom) to show that people who break problems into chunks, iterate, verify outputs, and bake checks into automated workflows get disproportionate value from models, while those who give up at the first hallucination or treat the model as an oracle complain it “doesn’t work.” Key technical implications: success often comes from prompt engineering, decomposition, verification layers, guardrails against hallucinations, and increasingly from orchestrating agentic systems that add limited autonomy but still require human direction. That has real consequences for the AI/ML community and industry: the technology amplifies existing differences in problem-solving skill, creating an adoption and capability divide with economic and information risks. Hallucinations, “AI slop,” and low-quality content hit hardest when users lack safeguards or critical workflows, enabling misinformation and displacement. Conversely, those who master tooling become exponentially more productive. The takeaway for practitioners: invest in human-in-the-loop design, testing/verification pipelines, and user training — because the question isn’t whether AI can get it done, but whether the user can shape it to.