AI Assisted Coding and Tools

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

Latest Discussions

Experience Report4d ago

AI Workflow & Tool Selection

User details a tiered LLM strategy: using Claude Opus for professional code maintenance and Sonnet for personal projects to manage token costs. The 'Dark Software Factory' concept involves using AI to build diverse, low-level projects (e.g., BASIC32 interpreters), illustrating how specific model strengths (like ChatGPT/Sol for planning) can optimize token efficiency and workflow.

Join the thread - 3 comments

🔗 Ask HN: What default model do you use and why?

General Discussion8d ago

Comparison of AI Workflow capabilities

A user inquiry seeking differentiation between a specific product (likely Shaurya) and Claude's dynamic workflows. The thread is currently truncated and contains a flagged comment, offering little actionable insight beyond identifying a competitive curiosity regarding AI workflow automation.

Join the thread - 2 comments

🔗 Show HN: Nightshift – Rust CLI to Orchestrate GitHub Issue Resolution with Dags

Innovation Discussion22h ago

Voice-activated AI agent workflows

Users are exploring hands-free, asynchronous AI coding tasks via mobile shortcuts. A key friction point identified is the friction caused by model safety/approval prompts, with users seeking 'hacky' automation methods to bypass manual approvals during autonomous workflows.

Join the thread - 2 comments

🔗 Show HN: AutoBot – live voice control for long-running AI work

Experience Report5d ago

AI-Assisted Coding Workflows

User reports a high-efficiency 3-step AI workflow (Discuss -> Task -> Implement) with a 50% acceptance rate. Key insights: AI is excellent for rapid implementation but carries risks of code duplication, logic errors, and 'degenerate codebases' if fully automated. There is a consensus that while AI tools accelerate development, they require heavy human oversight to prevent architectural drift.

Join the thread - 3 comments

🔗 How to Build an AI Software Factory: Agents That Open, Review, and Merge PRs

Technical Deep Dive2d ago

LLM Tool Integration & MCP

Users discuss the gap between developer-facing function calling and consumer-facing tool accessibility. The primary frustration lies in the lack of seamless integration between LLM voice/text modes and third-party ecosystems (MCP, GitHub, Notion) within mainstream consumer apps like Gemini.

Join the thread - 3 comments

🔗 Gemini 3.8 Live and 3.8 Live Extended Thinking

Innovation Discussion5h ago

Real-time collaborative AI coding vs. PR workflows

The discussion compares real-time collaborative/pair programming with traditional asynchronous Pull Request workflows. A key insight is that 'agentic coding' combined with live pair programming preserves context and prevents the 'context loss' and 'AI slop' common in traditional code reviews, potentially speeding up the merge cycle.

Join the thread - 2 comments

🔗 Replacing Pull Requests with Delta