AI Assisted Coding and Tools

Real-world experiences testing AI assisted coding and adopting AI tools into software development from Hacker News

Latest Discussions

Product Feedback4d ago

Cursor vs. Agentic CLI Tools

Users are debating the longevity of Cursor against emerging agentic CLI tools like Claude Code and Codex. Key differentiators for Cursor include its superior GUI/UX (visual diffs and agent windows), best-in-class autocomplete, and enterprise-grade features like Bugbot. However, competitors win on cost-efficiency (token subsidies) and performance (Zed/native IDEs). Actionable insights: Cursor's moat lies in its 'IDE-integrated' workflow vs. the 'terminal-heavy' agentic workflow, but it faces high pressure from model providers bundling their own interfaces.

Join the thread - 51 comments

🔗 SpaceX Is Buying Cursor

Product Feedback3d ago

AI Coding Tool Comparisons & Friction

Users compare Cursor, Zed, and Claude Code based on UX, autocomplete quality, and stability. Key insights include a preference for Cursor's 'Plan/Agent' modes despite UI quirks, friction with Zed's onboarding/bot detection, and the cost dilemma between subscription models vs. expensive API usage.

Join the thread - 6 comments

🔗 SpaceX to buy Cursor AI coding agent operator Anysphere for $60B

Experience Report6d ago

Practical AI Coding vs. Vibe Coding Workflows

Developers are increasingly using a mix of local models (via Ollama) and API tiers to perform targeted tasks like boilerplate reduction and debugging, rather than relying solely on frontier models. This approach avoids the need for expensive high-end hardware and prevents 'vibe coding' (generating low-quality code) by focusing on specific, high-utility AI-assisted workflows.

Join the thread - 5 comments

🔗 AI Coding at Home Without Going Broke

Experience Report6d ago

Evaluating AI Generation via Specific Prompts

Users are comparing AI model performance using specific test cases like the 'pelican' prompt and KingBench v3 benchmarks. Key takeaway: When evaluating model quality, focus on generation latency and the accuracy of fine structural details (e.g., bike frames).

Join the thread - 2 comments

🔗 GLM 5.2 Is Out

Technical Deep Dive3d ago

Multi-model AI coding workflows

Users are optimizing AI coding workflows by leveraging multi-model strategies to mitigate model blind spots. Key insights include using expensive models for high-level planning and cheaper models for execution, as well as utilizing CLI-based tools (like Cursor CLI) to maintain a lightweight, non-IDE-centric development environment.

Join the thread - 2 comments

🔗 SpaceX to buy Cursor AI coding agent operator Anysphere for $60B

Experience Report6d ago

Efficiency and Cost of AI Coding Tools

Users report that mid-to-low tier AI coding subscriptions ($20-$100/mo) are sufficient for most home use cases. A key takeaway is that hitting usage limits can be managed productively by using downtime for manual code review, which serves as a learning tool for improving prompting skills and understanding agent-generated logic.

Join the thread - 2 comments

🔗 AI Coding at Home Without Going Broke