AI Tools That Improved My Coding Workflow
August 24, 2025
I write software for a living, and I have spent the last year weaving AI into the parts of my workflow where it really helps. Debugging, code generation, architecture planning, and website scaffolding. Below is what has stuck, what has not, and why. I am not trying to be exhaustive or hypey, just honest about what improved my speed and focus.
How I judge tools, quick context
- I optimize for productive flow, fewer context switches, faster feedback loops, and clear diffs.
- I prefer portability over lock in, I like owning my infra and wiring third party services à la carte.
- I value explainability, why this change matters as much as what changed.
AI agents
Claude Code
https://www.anthropic.com/claude-code
Used, paid, 20 USD per month.
- I use it for multi file refactors, test generation, and architecture proposals.
- It explains intent before edits, I keep most of its patches with light tweaks.
- Daily limits are frustrating even on paid, I plan work in batches to avoid hitting the cap.
- My first choice when I want careful repo wide edits.
Qwen Code
https://github.com/QwenLM/qwen-code
Used, free plan.
- Works best when my prompt is very detailed with inputs, outputs, and acceptance criteria.
- I use it for utilities and scaffolds, I review carefully for security sensitive paths.
- Open source options make it easy to run locally or try hosted variants.
GitHub Copilot
https://github.com/features/copilot
Used, paid, 10 USD per month.
- Excellent inline completion, I save time on boilerplate and test skeletons.
- Copilot Chat helps with quick micro debugging inside the editor.
- I combine it with Claude Code for larger changes and review.
Cursor
Used, free plan, Pro available.
- The agent inside the IDE keeps me in one window and applies multi file edits on request.
- I still read every diff line by line before committing.
Tabnine
Haven’t used, paid.
- Research only in my notes, privacy friendly options and team tiers.
- Worth a trial if you want local or private completions.
Sourcegraph
Haven’t used, paid.
- Research only in my notes, strong repo context and enterprise controls.
- A good fit if your main pain is whole codebase search and answers.
Replit AI
Haven’t used, free plan available.
- Research only in my notes, idea to deployment flow inside the Replit ecosystem.
- Good when you want a cloud IDE plus AI in one place.
AskCodi
Haven’t used, paid.
- Research only in my notes, focuses on multi language assistance and snippets.
- Consider if you want a lightweight code helper outside the big players.
Snyk
Haven’t used, free and paid.
- Research only in my notes, security scanning for code and dependencies.
- Pairs well with any AI generator to catch unsafe patterns.
AI chats
ChatGPT
Used, paid, 20 USD per month.
- My default for broad problem solving, design trade offs, and quick example code.
- Great for rubber duck debugging and planning before I touch code.
- I verify outputs for deprecated APIs and version mismatches.
Claude.ai
Used, paid under the same Claude Pro plan.
- I use it for long context reviews and architecture discussions.
- Limits exist, I schedule sessions so a long thread does not stop mid way.
Gemini
Used, free plan.
- Handy for quick checks and a second opinion on designs and APIs.
- I keep it in the mix when I want another take without using paid quota.
Qwen Chat
Used, free.
- Solid for straightforward coding questions and doc analysis.
- I use it to cross check answers from other models.
Perplexity
Used, free plan.
- My pick when I want sources with the answer.
- Useful for API changes, RFCs, and research oriented debugging.
Website builders
Bolt
Used, free plan available.
- Fastest way I scaffold a full stack app, prompt to running project in minutes.
- I export the code and harden it when I need custom infrastructure.
- My personal favorite for speed to first deploy.
Lovable
Used, free plan available.
- Good for full stack prototypes with one click deploy and GitHub sync.
- The credit model encourages concise prompts and clearer specs.
v0
Used, free plan available.
- Great for UI iteration and design mode before I lock in component choices.
- Tight integration with Vercel helps me deploy quickly.
Anima
Used, free plan available.
- I use it for Figma to code handoff and fast design inspection.
- It speeds up design to code translation, I refactor generated code for maintainability.
Code review
GitHub Copilot
https://github.com/features/copilot
Used, paid, 10 USD per month.
- I run Copilot Chat on every medium or larger pull request to surface obvious issues.
- It drafts tests and summarizes diffs, I keep human review for complex logic.
Claude Code
https://www.anthropic.com/claude-code
Used, paid, 20 USD per month.
- I paste diffs or PR context and ask for risks and safer refactors.
- It is excellent at explaining intent changes across files.
ChatGPT Codex
Used as part of ChatGPT Plus.
- I use it for snippet audits and security focused questions on small diffs.
- My main review loop is Copilot and Claude Code.
Qodo
Haven’t used, free plan available.
- Research only in my notes, multi agent reviews and unit test suggestions.
- Worth exploring if you want automated PR comments.
Jules
Haven’t used, free plan available.
- Research only in my notes, asynchronous coding agent that files pull requests.
- Interesting option for repetitive maintenance tasks.
Backend as a service and AI
Supabase
Used, free plan available.
- I use it for Postgres, auth, and storage without heavy boilerplate.
- I prefer custom backend pieces with third party services, this keeps me flexible and reduces vendor lock in.
Database Build
Used, free.
- In browser Postgres with AI assistance, I scaffold schemas and test queries quickly.
- Helpful for data modeling conversations before I write migrations.
How these tools changed my workflow
- For debugging, I use ChatGPT and Claude to find the cause fast. They help me turn logs into small tests. I can review a first fix in minutes.
- For code generation, Copilot and Qwen Code write the basic parts and test stubs. I focus on the core logic and edge cases.
- For architecture planning, I test ideas in AI chats before I choose a path. I compare options, pick a simple design, and avoid rework.
- For website scaffolding, Bolt, Lovable, and v0 make a working preview from a prompt. I export the code, improve it, and fit it to my stack.
- For code review, Claude Code and Copilot summarize changes and highlight risks. I make the final call.
- For backend setup, Supabase gives me Postgres, auth, and storage with little setup. Database.build helps me design tables and try queries before I write migrations.