← Back

The Slop Tax: Action Plan

2026-03-12 #ai#engineering#productivity#product-thinking

  • AI amplifies everything. Including weakness.
  • A weak employee used to sit quietly. Now they generate slop that ends up in the product.
  • Removing slop from a product costs more than building from scratch.
  • Organizations measure lines of code, tasks, hours in Jira. They don’t measure: what the user actually got.
  • Those who generate volume have no answer to the question “why.”
  • Knowing how to write code has depreciated. Knowing why has become critical.
  • Coding is a solved problem. Claude Code, Cursor, Copilot write faster than most.

Developers: Learn Product Thinking

  • Three questions before the first line of code:
    1. What user problem are we solving?
    2. How do they solve it now?
    3. How will we know we solved it better?
  • Without answers — any prompt to an agent produces slop.
  • Add an "Export" button — slop.
  • User wants to export applications for a period to send to a contractor, currently copies by hand. Add CSV export with date and status filters — solution.
  • The difference isn’t in prompt length, but in thinking before the prompt.
  • Study Jobs To Be Done — user motivation.
  • Study User Story Mapping — scenario decomposition.
  • Study Impact Mapping — connecting features to business goals.
  • Write specs before code: who, what, why, how to verify.

Product Managers and Designers: Learn the Technical Side

  • A PM formulates the task but can’t verify the solution.
  • Doesn’t see where the architecture is falling apart.
  • Accepts slop as a result because “it seems to work.”
  • You don’t need to code — you need to understand how the system works.
  • Claude Code or Cursor — set tasks for agents at the code level.
  • Architecture basics: API, database, queue, separation of concerns.
  • Git — read diffs, understand what changed.
  • Formulate tasks so the agent doesn’t fill in the blanks.

The Skill of the Future: Product Thinking + AI Agent Systems Design

  • Product thinking answers: “what’s valuable.”
  • Systems design answers: “how to implement it.”
  • Don’t understand “why” — you generate slop.
  • Don’t understand “how to build” — you can’t verify the result.
  • Neither a pure developer nor a pure PM covers both sides.
  • The combination of product thinking + AI agent systems design is the skill of the future.

Conclusions

  • The cost of generating code trends toward zero.
  • The cost of understanding “why” and “how to verify” keeps growing.