Speed without a paper trail is just debt with good PR.
The Lazlo paradox
From the outside, Lazlo looks like a vibe coder. Same speed, same AI generation, same one-session builds.
But Lazlo is the most disciplined developer in the room.
And he'd genuinely love to show you why that's not a contradiction.
What vibe coding gets wrong
Vibe coding is real. The ability to build fast with AI is real. The problem isn't the speed.
The problem is that most vibe-coded projects have no spec, no contract tests, no documented decisions, and no way to be picked up cold by a fresh contributor — human or AI — six weeks later.
Fast to build. Slow to maintain. Impossible to hand off. Expensive to change.
That's not a coding problem. It's a documentation-and-architecture problem. And it's exactly what old-school engineering disciplines were designed to solve — before humans got too expensive, too tired, and too rushed to apply them consistently.
AI removes those constraints.
What Lazlo reconsiders
Lazlo asks a specific question about every methodology that got abandoned:
Were these bad ideas — or were humans simply too expensive, too tired, too rushed, and too inconsistent to do them right?
Waterfall planning died because builds took six months. AI builds in minutes. Deep upfront documentation died because diagrams went stale. AI maintains both code and diagram. Per-decision documentation died because humans got fatigued writing it. AI doesn't.
Lazlo isn't nostalgic. He's not bringing these back as costume. He's bringing them back because the constraint that killed them is gone.
How Lazlo builds
Capabilities — dumb black boxes. Each does one thing with clean inputs and outputs. Pure logic. The durable artifacts.
Wrappers — thin adapters that expose capabilities to specific consumers. CLI, REST, MCP. Same capability, multiple surfaces. Cross-cutting rules live here, not inside the capability.
Orchestrators — what decides which capabilities to call, when, in what sequence. Human-written saga code or AI reasoning over available tools.
Versioned Context — the source of truth for what the project is, what it's optimizing for, what rules must hold.
The paper trail
Every build produces three documentation tiers:
Architectural tier — diagrams of capabilities and key orchestration flows. The contract. Approved before any code is written.
Implementation tier — inline documentation, per-capability decision logs, data flow notes. Written for a naive future reader — human or AI.
Versioned Context — the intent document. What the project is, what it must never do, what rules apply everywhere.
A codebase is done when you can reconstruct every decision from the artifacts alone.
Who this is for
Founders and operators building internal tools, automations, and AI-orchestrated systems — who want them to actually work six months from now. People who've built something fast and are now afraid to touch it.