Historical source material recovered on 2026-09-09. Capture dates and source limitations apply; recovery is not fresh verification.

Specification As Coding

Specification as coding is the idea that future software work centers on writing the right level of intent, constraint, and direction for an AI or human collaborator, rather than always writing implementation code directly.

Key points

• Jensen says writing a specification is coding, and the artistry is choosing how specific or underspecified to be for the problem and collaborator ^[src-065].
• Prescriptive specification fits cases where the desired outcome is precise; underspecification fits exploratory work where AI or a large team can improve the idea ^[src-065].
• This reframes programming skill toward judgment, architecture, taste, problem decomposition, and calibration of autonomy ^[src-065].
• The concept connects to work anxiety: if someone's job is only a task, AI may disrupt it; if the job has a broader purpose, the person can use AI to automate tasks and elevate their role ^[src-065].

Related entities

• Jensen Huang
• Nvidia

Related concepts

• Coding Democratization
• Agentic Engineering
• Understanding Bottleneck
• AI Era Career Modernity
• AI Fluency As Language

Source references

• ^[src-065] Lex Fridman – "Jensen Huang: NVIDIA – The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494" (2026-03-23)

Historical recovered snapshot — 2026-08-23

Preserved from ai-watch-src393-ai-native-sdlc-playbook-20260823. Documentary recovery on 2026-09-09; claims below retain their historical source scope and were not freshly verified. Current synthesis above remains authoritative.

• Jensen says writing a specification is coding, and the artistry is choosing how specific or underspecified to be for the problem and collaborator ^[src-065].
• Prescriptive specification fits cases where the desired outcome is precise; underspecification fits exploratory work where AI or a large team can improve the idea ^[src-065].
• This reframes programming skill toward judgment, architecture, taste, problem decomposition, and calibration of autonomy ^[src-065].
• The concept connects to work anxiety: if someone's job is only a task, AI may disrupt it; if the job has a broader purpose, the person can use AI to automate tasks and elevate their role ^[src-065].
• Anthropic operationalises the idea as a chain of version-controlled intent.md, spec.md and approved plan.md artifacts before agent implementation begins ^[src-393].
• The specification is both delegation context and a review boundary: humans decide whether intent and trade-offs are acceptable before code generation, while tests and review evidence determine whether implementation matches the specification ^[src-393].

• ^[src-065] Lex Fridman – "Jensen Huang: NVIDIA – The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494" (2026-03-23)
• ^[src-393] Louis Claxton / Claude by Anthropic — “The AI-Native SDLC playbook” (2026-08-21)