{"slug": "how-i-get-better-results-from-claude-code", "title": "How I Get Better Results from Claude Code", "summary": "A developer at NRI Dating detailed a workflow for getting better results from Claude Code, arguing that process and context matter more than clever prompts. The practices include documenting project conventions in a repository instruction file, having the assistant inspect existing code before making changes, describing desired user behavior rather than implementation details, breaking large features into stages, and supplying screenshots for frontend work. The developer said asking the tool to inspect before editing has saved the most frustration, since AI assistants can write wrong code as quickly as correct code.", "body_md": "AI coding tools can dramatically speed up development, but only when you give them enough context and structure.\n\nWhile working on web and mobile development at [NRI Dating](https://nri.dating), I’ve increasingly used Claude Code as part of my development workflow. The biggest improvement didn’t come from learning clever prompts. It came from creating a better process around the tool.\n\nHere are the practices that have made the biggest difference for me.\n\n**Give It the Bigger Picture**\n\nOne of the easiest mistakes is asking an AI coding assistant to change something without explaining how that feature fits into the rest of the application.\n\nInstead of saying:\n\nAdd a block button to the profile page.\n\nExplain the actual behaviour:\n\nAdd a block option to the profile page. Once blocked, neither user should be able to access the other's profile. Existing conversations should remain visible but become read-only, and the profile link inside the conversation should no longer be accessible.\n\nThat extra context can prevent several rounds of fixes later.\n\nThe AI is no longer just being told what interface to create. It understands what the feature is supposed to mean across the product.\n\n**Keep Project Instructions in the Repository**\n\nFor larger projects, repeating your technology choices and conventions in every conversation becomes tedious.\n\nKeep a project instruction file containing things such as:\n\nIf your project uses Bootstrap and does not use Tailwind, say that once in the project instructions.\n\nIf there are existing patterns for controllers, services, API responses or database naming, document them.\n\nThe better the project context, the less time you spend correcting unnecessary changes.\n\n**Ask It to Inspect Before Changing**\n\nThis has probably saved me more frustration than anything else.\n\nBefore implementing a substantial feature, ask Claude Code to inspect the existing implementation first.\n\nFor example:\n\nInspect the existing authentication implementation, routes, middleware and database structure. Do not modify anything yet. Explain how it currently works and identify what needs to change.\n\nOnce you agree with its understanding, move on to implementation.\n\nAI coding assistants can write code incredibly quickly.\n\nUnfortunately, that also means they can write the wrong code incredibly quickly.\n\nUnderstanding first and editing second is usually much safer.\n\n**Describe the Result, Not Just the Code**\n\nYou do not necessarily need to know exactly how something should be implemented.\n\nDescribe what you expect the user to experience.\n\nA visitor can see the first page of search results. If they try to access additional results without having purchased coins, show the purchase prompt. Users who have previously purchased any coin package should continue normally.\n\nThat is often more useful than trying to dictate controllers, methods and database queries yourself.\n\nLet the coding assistant investigate the architecture and determine how the behaviour fits into it.\n\n**Break Large Features Into Stages**\n\nGiving an AI a massive specification and saying “implement everything” can work.\n\nIt can also create a spectacular mess.\n\nFor substantial features, I prefer something closer to:\n\nEach stage gives you an opportunity to catch a misunderstanding before it spreads through the project.\n\n**Tell It What Not to Do**\n\nConstraints are surprisingly useful.\n\nIf you do not want new dependencies:\n\nDo not install additional packages unless absolutely necessary.\n\nIf database compatibility matters:\n\nDo not rename or remove existing columns.\n\nIf you are fixing a production issue:\n\nMake the smallest change necessary to fix this problem. Do not refactor unrelated code.\n\nThese instructions reduce the tendency for a small task to turn into an unnecessary rewrite.\n\n**Screenshots Are Extremely Useful**\n\nFor frontend work, describing a visual problem can sometimes take longer than showing it.\n\nA screenshot plus:\n\nThe spacing between these cards is inconsistent. Keep the existing design but align the cards and make their vertical spacing consistent.\n\ncan communicate the requirement much more clearly.\n\nThe same applies when recreating a design.\n\nGive the assistant the visual reference and then explain which elements actually matter.\n\n**Make It Test Its Own Work**\n\nDo not stop at:\n\nImplement this feature.\n\nFinish with something like:\n\nAfter implementation, test the complete flow. Check validation, authorization, error states and existing functionality that could have been affected.\n\nFor bugs, ask it to reproduce the problem first whenever practical.\n\nThat encourages the assistant to verify that it fixed the actual issue rather than simply producing code that looks plausible.\n\n**Keep Changes Reversible**\n\nAI-assisted development makes changes extremely fast.\n\nThat makes reversibility even more important.\n\nBefore significant changes, make sure you have a reliable way to return to the previous state.\n\nThat might mean:\n\nThe best AI coding session is one where you can confidently experiment because nothing is irreversible.\n\n**Do Not Accept Everything It Suggests**\n\nClaude Code can often find solutions you would not immediately think of.\n\nIt can also confidently suggest unnecessary complexity.\n\nWhen it proposes:\n\nask why.\n\nSometimes the answer is excellent.\n\nSometimes you discover that the simpler existing implementation was perfectly adequate.\n\n**Give It Clear Ownership of the Task**\n\nOne thing that improves results is being explicit about what the assistant is responsible for.\n\nInstead of:\n\nCan you help me fix this?\n\nTry:\n\nInvestigate the issue, identify the root cause, implement the smallest safe fix, test it and report exactly what changed.\n\nThat creates a clearer workflow.\n\nIt also helps prevent situations where the assistant only explains the problem without actually finishing the task.\n\n**What I’ve Learned**\n\nMy biggest lesson from working with AI coding tools is surprisingly simple:\n\n**Do not spend all your time trying to write the perfect prompt. Build a workflow that gives the AI the right context.**\n\nExplain the outcome.\n\nLet it inspect the project.\n\nGive it constraints.\n\nBreak complicated work into stages.\n\nMake it verify its work.\n\nAnd keep yourself in control of architectural decisions.\n\nUsed that way, Claude Code is not just a faster way to type code. It becomes a genuinely useful development partner.", "url": "https://wpnews.pro/news/how-i-get-better-results-from-claude-code", "canonical_source": "https://dev.to/zarasethi/how-i-get-better-results-from-claude-code-11g0", "published_at": "2026-09-18 15:09:15+00:00", "updated_at": "2026-09-18 15:22:55.762899+00:00", "lang": "en", "topics": ["ai-tools", "ai-products", "developer-tools", "ai-agents"], "entities": ["Claude Code", "NRI Dating", "Anthropic", "Bootstrap", "Tailwind"], "alternates": {"html": "https://wpnews.pro/news/how-i-get-better-results-from-claude-code", "markdown": "https://wpnews.pro/news/how-i-get-better-results-from-claude-code.md", "text": "https://wpnews.pro/news/how-i-get-better-results-from-claude-code.txt", "jsonld": "https://wpnews.pro/news/how-i-get-better-results-from-claude-code.jsonld"}}