Vibe Coding: An Author's Agent Dev Workflow
数字生命卡兹克 · wechat · 2026-07-15
The author shares their Vibe Coding workflow from the past two weeks: spending long hours daily collaborating with Claude, GPT-5.6, and Codex. They emphasize that the primary bottleneck has shifted from "writing code" to "testing, validation, and design review."
Core Workflow
- Use Claude for research and initial design drafts.
- Pass the requirements and design to GPT-5.6 for review, error correction, and optimization.
- Finally, activate "Goal Mode" in Codex, allowing the Agent to execute continuously until the objective is met.
Key Takeaways
- As projects scale, CI and automated testing become critical, often consuming more time than writing code itself.
- The author set up a cheap CI server on Tencent Cloud to reduce GitHub hosted Runner wait times.
- Codex's "Fast Mode" isn't necessarily faster for large projects, as time is heavily consumed by deterministic testing pipelines.
- The most effective approach isn't a clever trick, but closing the loop between "design, automated execution, rigorous testing, and documentation/memory sync."
Practical Example
Using their own AI trending product as an example, the author illustrates how to let Agents automatically open PRs, run tests, fix bugs, re-submit for testing, and finally deploy and review. They believe the best current strategy is to have the strongest models handle design and error correction, let Agents execute for extended periods, and reserve human intervention for final confirmation.
Related event: Coding Agents Shift Toward Orchestration and Verification(6 posts)→
More from coding & agent
- Alex Townsend posts 200 open problems in numerical linear algebra for humans and AI agents — IgorCarron · 2026-09-11
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11
- Looking for a classifier of software engineering task shapes to pick models per task — StewartalsopIII · 2026-09-11
- Steal this idea: prompt-to-hardware where agents assemble custom devices — paraschopra · 2026-09-11
- Model Is the Least Interesting Part: A Guide to Six Core AI Architectures from RAG to Multi-Agent — goyalshaliniuk · 2026-09-11
- Non-coder builds layered memory architecture: 20k tokens tracks a year of agent conversations — matteoianni · 2026-09-11