AI may write your code. You own the consequences.
The newest tool · The oldest responsibility
Kode-Fu and Vibe Coding
AI makes code dramatically easier to produce. That is an opportunity: as creation accelerates, understanding and engineering judgment become more valuable—not less.
01 // The New Apprentice
The apprentice arrived with infinite patience and astonishing recall.
It can sketch an interface, explain an unfamiliar library, propose a migration, and turn an idea into running code before the coffee cools. Used well, an AI coding assistant removes friction, widens access, and gives experienced engineers more room to think at the level of systems.
Vibe coding can be an energetic way to explore: describe an intention, inspect what appears, and iterate toward something useful. Exploration is real work. The important question is what happens when an experiment becomes a dependency.
02 // AI Is the Newest Tool
Tools evolve.
Principles endure.
Compilers automated machine instructions. Frameworks automated repeated architecture. Cloud platforms automated infrastructure. AI now automates a growing share of expression. Each step changes what engineers type and raises the level at which they can work.
None of these tools can choose which consequence your team is willing to own. They do not know your users, operating constraints, institutional history, or acceptable risk unless you make that context explicit—and even then, the decision remains yours.
03 // Where Kode-Fu Begins
Anyone can generate code.
Kode-Fu begins when that code must survive production.
Production is where plausible output meets real data, real latency, real attackers, real maintenance, and real people. The move from generation to stewardship is the threshold.
04 // The Five Gates
Before generated code enters production
The gates are questions, not bureaucracy. Their depth should match the risk, but none should be skipped by accident.
- 01
Understanding
Can you explain what the generated code does, why it is shaped this way, and which assumptions it makes?
- 02
Correctness
Does it meet the real requirement across edge cases, failure modes, permissions, and unexpected input?
- 03
Maintainability
Can the next engineer change it safely without reconstructing the conversation that produced it?
- 04
Reliability
Can it be observed, operated, recovered, and trusted when production behaves differently from the prompt?
- 05
Responsibility
Who owns its impact on users, security, cost, data, and the people asked to support it?
05 // Ancient Wisdom
New capability, familiar disciplines
The metaphors are educational bridges. They are not claims that software engineering and historical traditions are equivalent.
06 // Timeless Principles in the Age of AI
Judgment is the force multiplier
Models generate possibilities. Engineers accept responsibility.
Fast output increases the value of careful verification.
A prompt can start the work. Shared understanding must finish it.
07 // Final Teaching
Never has software been easier to write.
Never has engineering judgment been more valuable.
Powerful tools do not diminish the craft. They give us more leverage—and ask us to apply that leverage with clarity, evidence, humility, and care.