💡 Use Cases
Real-world examples and success stories
Built a Daily YouTube → LinkedIn Pipeline with OpenClaw: Architecture, Gotchas, and Lessons Learned
A developer shares the architecture of an OpenClaw skill that scrapes 30 YouTube channels daily, analyzes transcripts via LLM, and writes to Google Sheets. Details key gotchas like Apify async vs sync, Codex idle-turn watchdog, and ARG_MAX limits.
Claude Code Audits 80-Component React Library Docs: Real Bugs Found, New Bug Introduced
A staff engineer used Claude Code to audit docs for an 80-component React library. It caught real bugs but also introduced new ones requiring a review pass.
Inside the $20.8K MRR Feature: 60 Prompts Over 14 Months on Claude
A tutoring platform built a session summary feature with Claude in 3 hours, then refined the prompt 60+ times over 14 months. The feature drives 22% of parent conversions and contributes to $20.8K MRR.
Practical Criticism of LLM Memory: Immutable Reflections & Ephemeral Sessions as Solutions
A critique of long-running sessions, life-companions, and LLM-wiki memory, offering solutions like immutable reflections, issue-bound session chains, and prompt templates to avoid intent loss and context overload.
Master Thesis Written Mostly with Claude: Economics Student Passes with Top Grade
An economics student used Claude for literature review, data analysis with Python scripts and Excel, and presentation — passed with a top grade near PhD level. Appendix disclosed AI use, no questions asked.
Building a Full App with Lovable + Claude + Gemini: A Case Study
A software engineer built Earnest, a bank bonus tracker, using Lovable for UI, Claude for intent understanding, and Gemini as a second opinion. The app now has 100+ users tracking $9,700+ in bonuses.
How One Team Replaced a 6-Figure HubSpot Agency with Claude Code
A mid-sized e-commerce company built their entire HubSpot Enterprise migration using Claude Code, replacing quotes of 20k-80k EUR for partial setups. They built 6 custom objects, 5 n8n integrations, and a KlickTipp migration in 4 months, with Claude Code handling both code and documentation.
Agentic Infrastructure: Replacing Splunk with Claude Code Agents for Server Monitoring
A developer deploys Claude Code sessions as services — router, monitors, dashboard poller — connected via WebSocket hub. Watchers are cheap bash; LLM wakes every 5 min for drain cycle. Dashboard tiles are natural-language queries cached in SQLite.
Enterprise AI agents: OpenClaw for channels, custom MCP tools, Cursor CLI runtime
Running AI agents in production for compliance, devops, and finance requires deterministic tooling, not raw API access. This post details a recipe: OpenClaw for channels, custom MCP per process, Cursor CLI as the agent runtime via ACPX, and self-hosted Kubernetes with immutable agent code.
Debugging a Pi Zero 2W BadUSB with Claude Code: Fixing an 'Impossible' Bug
A developer rebuilt a Pi Zero 2W BadUSB toolkit with Claude Code, which diagnosed a wrong-signal bug, empirically confirmed hardware limitations, and fixed a silent Python no-op in under 4 hours.
Claude for Engineering Compliance: 6-Month Workflow Breakdown
A technical firm shares how they use Claude Projects, Artifacts, and constraint-following to avoid hallucinations in client-facing specs.
Analyzing 7 Years of Diary Entries with an LLM: RAG vs Fine-Tuning Failures
After keeping a diary since 2019, a developer fed 200+ entries to an LLM to discover patterns — RAG failed, fine-tuning failed, and privacy was a constraint. The final approach revealed cyclical life lessons every two years.