Exploratory AI coding, or 'vibe coding,' proved catastrophic for production environments. The most effective developers adapted by treating AI like a junior engineer, providing lightweight specifications, tests, and guardrails to ensure the output was viable and reliable.

Snyk founder's new venture, TESOL, posits that AI will make code disposable. Instead of code being the source of truth, a durable, versioned 'spec' document defining requirements will become the core asset. AI agents will generate the implementation, fundamentally changing software development.

Cisco is developing its AI defense product entirely with AI-written code, with human engineers acting as "spec developers." This fundamentally changes the software development lifecycle, making code review—not code creation—the primary bottleneck and indicating a future where engineering productivity is redefined.

Cognition's Scott Wu predicts that AI will elevate software development to a new level of abstraction. Instead of reviewing code, engineers will review and iterate on English-language specifications and product decisions. The AI agent will handle the code generation, making English the new "source of truth."

Inspired by fully automated manufacturing, this approach mandates that no human ever writes or reviews code. AI agents handle the entire development lifecycle from spec to deployment, driven by the declining cost of tokens and increasingly capable models.

With autonomous AI coding loops, the most leveraged human activity is no longer writing code but meticulously crafting the initial Product Requirements Document (PRD) and user stories. Spending significant upfront time defining the 'what' and 'why' ensures the AI has a perfect blueprint, as the 'garbage-in, garbage-out' principle still applies.

Instead of writing code, engineers verbally describe a feature, use an AI to generate a detailed spec, and then point another AI agent at the spec to build the feature. The spec file becomes the source of truth, managed in version control.

As AI writes most of the code, the highest-leverage human activity will shift from reviewing pull requests to reviewing the AI's research and implementation plans. Collaborating on the plan provides a narrative journey of the upcoming changes, allowing for high-level course correction before hundreds of lines of bad code are ever generated.

A powerful technique for creating robust software plans is to use AI as an adversarial partner. After drafting a specification, prompt an AI to "tear it apart" by identifying underspecified or inconsistent points. Iterate on this process until the AI's feedback becomes niche, indicating a solid spec.