AI Coding Guide

AI-Assisted Software Engineering

Beyond Coding

By Richard Osborne, CTO at Visual Hive

Last updated:

APPENDIX

The core methodology — brainstorm first, document thoroughly, execute in focused tasks, maintain quality gates — applies to any complex knowledge work. Software development is just one application. This appendix explores others.

The Common Pattern

What makes the methodology work for software:

  1. A real conversation defines scope and approach before any execution
  2. Documentation carries context across sessions so AI starts informed
  3. Focused tasks with clear acceptance criteria prevent drift
  4. Quality gates catch problems before they compound

None of this is specific to code. Replace "task spec" with "brief" and "confidence score" with "editorial review" and you have a writing workflow. The structure transfers.

Content and Writing

The documentation equivalent for a content project:

  • Audience doc — who reads this, what do they already know, what tone
  • Structure doc — outline, section purposes, word counts
  • Voice guide — dos/don'ts, examples of good/bad writing for this project
  • Brief template — context, goals, specific instructions per piece

The brainstorming session defines the content strategy. Foundation docs capture editorial standards. Individual briefs are the task specs. Editorial review is the confidence score.

Research and Analysis

For a research project:

  • Brainstorm defines the research questions and methodology
  • A "Research Architecture" doc captures the framework, key sources, known constraints
  • Each research question gets a focused task session
  • A review prompt checks for logical gaps, unsupported claims, missing perspectives

The fresh-conversation principle is especially important in research — AI carrying the weight of prior analysis is more likely to find confirmation than new insight.

Strategy and Planning

The Opus brainstorming conversation is particularly valuable for strategic work. Push Claude to steelman opposing views, identify assumptions, and challenge the reasoning. The output isn't code — it's a strategy document with clear decision records.

The quality rule equivalent: "Don't give me the answer you think I want. Tell me what concerns you."

Design Briefs

For design projects, the methodology provides:

  • Brainstorm for design direction, constraints, success criteria
  • A design brief template that includes brand context, requirements, non-negotiables
  • Component-level briefs for each piece of work
  • Review sessions that check consistency against the brief, not just aesthetic quality

Operations and Processes

Process documentation, standard operating procedures, training materials — all benefit from documentation-first thinking. The "ARCHITECTURE.md" becomes a process map. The ".clinerules" becomes the quality standards document. Tasks become the specific processes being documented.

The Transferable Insight

The reason AI-assisted coding fails without structure is the same reason AI-assisted anything fails without structure: AI is very capable of execution, but it needs clear scope, quality standards, and sufficient context to do it well.

The methodology provides all three. Whether you're building software, writing a report, designing a product, or planning a campaign — the pattern works.

Building something with AI?

Talk to Visual Hive →