Taming AI Coding Agents: How Pattern-Based Codebases Using Effect Act as an Intermediate Representation
samgoodwin89 · x · 2026-07-23
Developer Sam Goodwin shared the engineering practice of using "pattern-based codebases" combined with the Effect framework at Alchemy.
The core insight is that AI coding agents often go off the rails when facing open-ended design questions. By establishing repeatable patterns and conventions, developers can constrain the AI, forcing it to extrapolate from existing structures rather than inventing new subsystems. The Effect framework makes this practical through built-in features like dependency injection, tagged errors, and services, which eliminate whole categories of design decisions—acting effectively as an intermediate representation (IR) for AI. This high level of consistency not only improves the accuracy of AI-generated code but also reduces cognitive load for human developers.
More from coding & agent
- Claude Code plus Pika’s Voxel-it skill turns a meetup photo into a Minecraft scene — ytjessie_ · 2026-07-23
- A $20 Ethernet link is enough for 39.7GB multi-node GPU inference — Chuyito · 2026-07-23
- OpenAI GPT-5.6 Sol helped solve 6 open Erdős problems in 5 days — Charuru · 2026-07-23
- ModelScope launches an MCP server for text-to-image generation with Qwen-Image — modelcontextprotocol · 2026-07-23
- AILANG Parse brings deterministic DOCX, PPTX, XLSX and PDF parsing to MCP — modelcontextprotocol · 2026-07-23
- Agent workflow pitch: one Chief of Staff can spin up VMs and centralize output — mattshumer_ · 2026-07-23