PromptWright: Natural Language to Self-Healing Playwright Tests
- QA veteran Amit Rawat built PromptWright, a desktop tool converting natural language prompts into Playwright tests with screen recording, Gherkin...

Created by Taylor Smith
First‑person demos of clever AI assistants, automation, and knowledge‑management workflows
Explore the latest content tracked by AI Productivity Playbook
A lightweight PKM system turns raw notes into connected wiki pages using Obsidian for navigation and Claude Code as the processing agent.
A journalist ditched Claude for Perplexity Computer on crunch days, keeping both tools for complementary roles.
These clever setups turn a smart home into a proactive productivity coach and daily-life assistant.
A first-person walkthrough reveals a privacy-focused local setup rivaling cloud assistants:
Sutha Kamal commands his Cortex AI chief-of-staff via T-Mobile satellite iMessage while hiking with zero cell service, dispatching coding agents,...
Apple Notes plus three Shortcuts beats Obsidian for most people by slashing capture friction and maintenance.
Production teams default to building their own agent runtime over third-party frameworks to avoid long-term maintenance burdens.
Framework options at...
A single AutoHotkey v2 script turns repetitive friction into instant action, delivering a night and day productivity shift.
A Walmart regional load manager built an AI app after completing the company's Google AI certification to automatically match truckers with nearby...
openHAB MCP Server lets MCP clients like Claude or ChatGPT read and control your smart home via simple natural language commands. Install the add-on,...
Two real-world examples show analytical AI tackling routine mental overhead beyond generative tools.
Plano routes prompts to cheaper models like gpt-4o-mini for routine tasks while reserving claude-sonnet-4.5 for complex work, delivering the reported...
Building your own ESP32 devices with Home Assistant delivers local control and expandability without cloud dependencies or app sprawl.
Jakub Jirák's fully local workflow uses Ollama + Qwen2.5 14B to deliver verdicts on ~50 daily RSS items in a 20-minute pass, keeping all data...
Three tools aim to wrap local models with memory, docs, tools, and agents, but differ sharply in delivery.