--- title: My AI Coding Workflow That Actually Works | daily.dev description: A practical AI-assisted development workflow built around structured agents, spec files, and roadmaps stored in a Git repository. The workflow emphasizes... canonical: https://daily.dev/posts/my-ai-coding-workflow-that-actually-works-enn37rhav twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: My AI Coding Workflow That Actually Works | daily.dev og:description: A practical AI-assisted development workflow built around structured agents, spec files, and roadmaps stored in a Git repository. The workflow emphasizes... og:url: https://daily.dev/posts/my-ai-coding-workflow-that-actually-works-enn37rhav og:image: https://api.daily.dev/og/posts/eNn37RHAV.png og:image:alt: My AI Coding Workflow That Actually Works og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # My AI Coding Workflow That Actually Works **[Chris Titus Tech](https://daily.dev/sources/christitustech)** · 25 min read · 1 upvotes · 0 comments ## Summary A practical AI-assisted development workflow built around structured agents, spec files, and roadmaps stored in a Git repository. The workflow emphasizes manual review gates at each phase rather than letting LLMs run unchecked. Key components include: writing spec files that pin dependency versions and define acceptance criteria, building a test suite before coding begins, using AI for code review (CodeRabbit, Codex review) alongside CI pipelines, and chunking work into small pull requests. The author demonstrates the workflow live by adding a volume percentage display to a QuickShell desktop environment, showing how AI handles implementation in minutes while human oversight catches edge cases like missing audio states. Cost considerations around $100-200/month AI subscriptions and the potential future shift to local models are also discussed. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can I stop AI coding agents from producing poor results on large software projects? Poor results usually come from skipping planning and just using a single prompt to implement a whole feature at once. A better approach uses three components: a machine-wide agents file defining coding rules and style, a repository-specific spec file describing the problem, architecture, and acceptance criteria, and a roadmap breaking work into phases. Build a testing suite and scaffolding first, then implement with a mandatory manual review gate at each phase via pull request before merging. _Developers refining their own AI coding workflow can compare structured approaches like this on daily.dev._ ### Why do LLMs often pick outdated or insecure dependency versions when writing code? Large language models tend to default to older package and library versions that have known security problems, since their training data skews toward what was common at the time rather than the latest releases. Pinning the exact version being used in a spec file, and enabling Dependabot and CodeQL code scanning on the repository, helps catch outdated or vulnerable dependencies automatically and for free. _Anyone hardening AI-generated code against stale dependencies can track these version-pinning practices on daily.dev._ ### How much does heavy AI coding usage cost on Codex subscription plans? Heavy development work with independent code reviews and iterative testing burns through Codex usage quickly; a single day of intensive review-heavy work consumed 25% of a $100/month max plan's quota after a fresh reset. A $20/month plan would be exhausted even faster under this kind of workflow, meaning serious daily development work effectively requires the $100-$200 subscription tiers. _Developers budgeting for AI coding subscriptions can weigh real usage costs like these on daily.dev._ ## Similar posts on daily.dev - [My AI Workflow \(Without Losing My Skills\)](https://daily.dev/posts/my-ai-workflow-without-losing-my-skills--yabivimt9) · Marc G Gauthier · 3 upvotes · 1 comments - [Automate Your Code Reviews: Welcome to the Post-AI Era of Software Development.](https://daily.dev/posts/automate-your-code-reviews-welcome-to-the-post-ai-era-of-software-development--9qd0ky8y4) · ITNEXT · 0 upvotes · 0 comments --- Tags: [#testing](https://daily.dev/tags/testing), [#git](https://daily.dev/tags/git), [#ai-coding](https://daily.dev/tags/ai-coding), [#code-review](https://daily.dev/tags/code-review) [View this post on daily.dev](https://daily.dev/posts/my-ai-coding-workflow-that-actually-works-enn37rhav) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"My AI Coding Workflow That Actually Works","url":"https://daily.dev/posts/my-ai-coding-workflow-that-actually-works-enn37rhav","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/my-ai-coding-workflow-that-actually-works-enn37rhav"},"datePublished":"2026-07-29T11:21:49.704Z","dateModified":"2026-09-13T19:05:29.570Z","description":"A practical AI-assisted development workflow built around structured agents, spec files, and roadmaps stored in a Git repository. The workflow emphasizes...","image":"https://i.ytimg.com/vi/wcRR5P0S2Us/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/wcRR5P0S2Us/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Chris Titus Tech","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Chris Titus Tech","logo":"https://media.daily.dev/image/upload/s--4e28ySkK--/f_auto/v1710504019/logos/christitustech","url":"https://daily.dev/sources/christitustech"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/my-ai-coding-workflow-that-actually-works-enn37rhav","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"testing,git,ai-coding,code-review","timeRequired":"PT25M","video":{"@type":"VideoObject","name":"My AI Coding Workflow That Actually Works","description":"A practical AI-assisted development workflow built around structured agents, spec files, and roadmaps stored in a Git repository. The workflow emphasizes...","thumbnailUrl":"https://i.ytimg.com/vi/wcRR5P0S2Us/sddefault.jpg","uploadDate":"2026-07-29T11:21:49.704Z","duration":"PT25M","url":"https://api.daily.dev/r/eNn37RHAV","embedUrl":"https://www.youtube.com/embed/wcRR5P0S2Us"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Chris Titus Tech","item":"https://daily.dev/sources/christitustech"},{"@type":"ListItem","position":3,"name":"My AI Coding Workflow That Actually Works"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/my-ai-coding-workflow-that-actually-works-enn37rhav#faq","mainEntity":[{"@type":"Question","name":"How can I stop AI coding agents from producing poor results on large software projects?","acceptedAnswer":{"@type":"Answer","text":"Poor results usually come from skipping planning and just using a single prompt to implement a whole feature at once. A better approach uses three components: a machine-wide agents file defining coding rules and style, a repository-specific spec file describing the problem, architecture, and acceptance criteria, and a roadmap breaking work into phases. Build a testing suite and scaffolding first, then implement with a mandatory manual review gate at each phase via pull request before merging. Developers refining their own AI coding workflow can compare structured approaches like this on daily.dev."}},{"@type":"Question","name":"Why do LLMs often pick outdated or insecure dependency versions when writing code?","acceptedAnswer":{"@type":"Answer","text":"Large language models tend to default to older package and library versions that have known security problems, since their training data skews toward what was common at the time rather than the latest releases. Pinning the exact version being used in a spec file, and enabling Dependabot and CodeQL code scanning on the repository, helps catch outdated or vulnerable dependencies automatically and for free. Anyone hardening AI-generated code against stale dependencies can track these version-pinning practices on daily.dev."}},{"@type":"Question","name":"How much does heavy AI coding usage cost on Codex subscription plans?","acceptedAnswer":{"@type":"Answer","text":"Heavy development work with independent code reviews and iterative testing burns through Codex usage quickly; a single day of intensive review-heavy work consumed 25% of a $100/month max plan's quota after a fresh reset. A $20/month plan would be exhausted even faster under this kind of workflow, meaning serious daily development work effectively requires the $100-$200 subscription tiers. Developers budgeting for AI coding subscriptions can weigh real usage costs like these on daily.dev."}}]} ```