Vinko Novak contrasts AI as search replacement with AI as delegation and argues that many teams confuse feeling faster with operating near the actual frontier of speed.
This article draws a line between two modes of AI-assisted development: asking a model questions inside a chat window, versus designing agents that execute repeatable work end-to-end.
Its central claim is that many developers overestimate their effectiveness because AI makes them feel faster than before, even though they are still far from the actual productivity frontier.
The practical recommendation is to start with a small repeatable task, document it clearly and let an agent run it with minimal human orchestration.
Related
An international testing and certification group set out to generate quotes for company certifications automatically. The solution went into production in 2018 – with a human in the loop and a weekly C-level steering committee.
Open detail pageVibe Coding is almost free to start – but the bill arrives later: token burn, maintenance tax, security cleanup. Beyond a crossover point, Agentic Engineering delivers 3–10× lower TCO per feature. The decisive lever is total cost of ownership, not startup cost.
Open detail pageA compact overview of around 50 slash-commands in Claude Code – what they do, when to use them, and how they accelerate the AI coding workflow.
Open detail page