Build the loop once. Let it run.
Loopwise tests AI tools and automation workflows, then publishes what works along with the code to run it today.
Latest articles
All articles- ChatGPT vs Claude: How to Choose for Your Own Work Skip the benchmark noise. Run a five-task, blind test in about an hour to decide between ChatGPT and Claude for your own writing, coding and research.
- Claude Tool Use Explained With Three Real Python Examples Build one reusable tool-use loop, then use it for an order lookup, a read-only SQL assistant and a calendar tool that asks a human before it acts.
- Draft Email Replies With AI and Never Let It Send Build a Python script that reads your inbox, drafts grounded replies with Claude, saves them as drafts for you to review, and defends against hostile emails.
- How to Get Reliable JSON From an LLM (With Python Code) Stop parsing broken JSON. Force a schema with tool use, validate with Pydantic, feed errors back for a retry, and measure how often your prompt really works.
- How to Use the Claude API in Python: A Beginner's Guide Install the SDK, make your first request, stream replies, hold a conversation, call your own functions and handle errors, with Python code you can run.
- Web Scraping with Python and AI: Turn Pages Into Clean Data Fetch pages politely, strip them to the text that matters, and have Claude return validated JSON. Includes robots.txt checks, caching and cost control.
Tools worth knowing
All toolsOne tested workflow a week
Each issue covers a single automation: what it does, the code or setup, and what it cost to run.