Bad code, progressive complexity, and weekly readings💡

2025-04-30 · Source: Refactoring · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

Salesforce's TDX event highlighted Agentforce as its fastest-growing product, introducing over 60 new MCP tools in Headless 360, open-sourcing Agent Script, and releasing Agentforce Vibes 2.0, alongside new Slack agent capabilities. The company also published the full language specification for its agent-definition language, signaling a strategic bet on an open agent ecosystem for developer adoption. Concurrently, the Tolaria project, an AI-generated codebase, garnered 6000+ GitHub stars in under a week, demonstrating that "bad code" (e.g., lack of tests, high complexity) can be largely mitigated by enforcing quality gates on AI output. However, "misaligned code" (clean but wrong for product direction) remains a human challenge. Guillermo Rauch, Vercel's CEO, advocates for "progressive disclosure of complexity" in technology design, ensuring products like Next.js are accessible to beginners while scaling for enterprise needs through "token minimization" and consistent infrastructure.

Key takeaway

For AI Engineers and Directors of AI/ML evaluating AI's role in code generation, you should focus AI tools on eliminating "bad code" by integrating strict quality gates into your CI/CD pipelines. This frees human developers to concentrate on critical architectural decisions and strategic product alignment, areas where AI currently falls short. Prioritize reviewing abstractions and future intent, as these are not solvable by AI alone.

Key insights

AI can largely solve "bad code" issues through enforced quality gates, but human oversight remains crucial for architectural alignment.

Principles

• Enforce quality gates for AI-generated code.
• Prioritize architectural review over basic code hygiene.
• Design technology for progressive complexity disclosure.

Method

Enforce AI-generated code quality via CI checks for test coverage and health scores; design APIs with "token minimization" for progressive complexity.

In practice

• Explore Salesforce's open-sourced Agent Script.
• Integrate AI code quality checks into CI.
• Review architecture for future product alignment.

Topics

• Salesforce Agentforce
• AI Code Generation
• Technical Debt
• Progressive Disclosure
• API Design
• CI/CD Quality Gates

Code references

• refactoringhq/tolaria

Best for: CTO, VP of Engineering/Data, AI Architect, AI Engineer, Software Engineer, Director of AI/ML

Related on AIssential

• AI Is Writing Our Code Faster Than We Can Verify It — AI can make traditional quality engineering affordable, enabling scalable verification of AI-generated code.
• AI should help us produce better code — AI coding agents can enhance code quality and reduce technical debt by automating refactoring and enabling rapid prototyping.
• AI Can Generate Code in Minutes. Reviewing It Responsibly Is the Real Engineering Skill — AI accelerates code generation, but shifts the engineering bottleneck to responsible review and validation.
• Are AI agents actually slowing us down? — Over-reliance on AI coding tools prioritizes speed over quality, leading to increased tech debt and system outages.
• PR reviews were already broken. AI made it worse — AI-coding agents exacerbate PR review issues by generating high-volume, low-quality code from insufficient data, demanding a layered verification approach.
• Code smells for AI agents: Q&A with Eno Reyes of Factory — High-quality codebases are crucial for AI agent productivity, as agents excel at pattern recognition.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Refactoring.