{"slug": "working-with-ai-what-actually-works-for-me", "title": "Working With AI: What Actually Works For Me", "summary": "A developer describes a workflow where AI agents handle implementation while humans focus on planning, architecture, and system design. The key insight is that the quality of AI-generated code depends on the quality of the plan, and that enforcing rules through deterministic mechanisms is more reliable than relying on instructions. The workflow involves breaking tasks into small pieces, using separate models for implementation and review, and maintaining a written architecture document.", "body_md": "I think a lot of people still imagine AI coding as opening ChatGPT, asking for code, and copy-pasting the result. That's not really how I work anymore.\n\nThe biggest shift for me is that planning matters far more than coding.\n\nEarlier, execution was expensive, so most of the effort went into writing code. Now execution is cheap. I can have an agent implement something in minutes. The hard part is making sure the plan is correct.\n\nMost of my effort goes into thinking through the architecture, edge cases, failure modes, test strategy, and how the change fits into the broader system. If the plan is vague, the agent will confidently implement the wrong thing. The quality of the result is mostly determined by the quality of the plan.\n\nOnce I have a plan, I break it into small independent pieces. Each piece should be executable without additional clarification. If an agent needs to stop and ask questions, the task probably isn't broken down enough.\n\nThose pieces become tickets.\n\nThen an agent picks up a ticket and implements it.\n\nThe important thing is that the agent isn't operating in a vacuum. I try to give it a good environment to work in:\n\nOne lesson that really stuck with me is that instructions are guidance, not guarantees.\n\nAt one point I had \"always use a git worktree\" written in AGENTS.md. The model still ignored it occasionally. When I dug into it, the answer was simple: models can drift from instructions.\n\nSo if something absolutely must happen, don't rely on instructions. Enforce it. Put it in a hook, script, validation step, CI check, or some other deterministic mechanism.\n\nIf it is important, make it impossible to skip.\n\nOnce the implementation is done, the agent opens a PR.\n\nThis is where another useful pattern comes in: don't let the same model review the code it wrote.\n\nI usually have one model implement and another model review.\n\nDifferent models catch different things. They have different strengths, weaknesses, and blind spots. The disagreements between them are often where the real issues are hiding.\n\nA surprising amount of feedback can be handled at this stage before a human ever looks at the PR.\n\nAfter that, human review depends on what changed.\n\nIf the change touches critical infrastructure, security, core architecture, data correctness, or anything with a large blast radius, I want human eyes on it.\n\nFor lower-risk changes, I care about two things:\n\nIf both are true, I don't think every change needs a mandatory human line-by-line review.\n\nThe other thing that has become important is having a written constitution for the codebase.\n\nFor me this is usually an architecture document that defines the rules of the system.\n\nFor example, if the architecture says inner layers cannot depend on outer layers, then that rule should be written down explicitly. Not tribal knowledge. Not something that only senior engineers know.\n\nWhen the rules are documented, every agent can follow them consistently.\n\nThe workflow ends up looking something like:\n\nPlan → Break into small tasks → Create tickets → Agent implements → Different model reviews → Human reviews critical changes → Merge\n\nWhat's interesting is how little of my time is spent writing implementation code.\n\nMost of my time goes into:\n\nAnd that's probably the biggest change.\n\nThe real leverage isn't getting AI to write code.\n\nThe real leverage is continuously improving the system around the AI:\n\nThe better the harness becomes, the better every future task becomes.\n\nThat's where most of the compounding value comes from.", "url": "https://wpnews.pro/news/working-with-ai-what-actually-works-for-me", "canonical_source": "https://dev.to/miteshethos/working-with-ai-what-actually-works-for-me-37nf", "published_at": "2026-06-16 15:36:21+00:00", "updated_at": "2026-06-16 15:47:17.224419+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "developer-tools", "ai-agents", "ai-products"], "entities": ["ChatGPT", "GitHub"], "alternates": {"html": "https://wpnews.pro/news/working-with-ai-what-actually-works-for-me", "markdown": "https://wpnews.pro/news/working-with-ai-what-actually-works-for-me.md", "text": "https://wpnews.pro/news/working-with-ai-what-actually-works-for-me.txt", "jsonld": "https://wpnews.pro/news/working-with-ai-what-actually-works-for-me.jsonld"}}