All TWiT.tv Shows (Audio)
Hands-On Tech 255: Bulk Recipe Imports
Host: Micah Sargent
Date: February 22, 2026

Episode Overview

In this episode of "Hands-On Tech," Micah Sargent addresses a listener’s question about streamlining the process of scanning and importing large batches of physical recipe cards into the Mela recipe manager app. Through practical advice and a deep dive into AI-tools, automation scripts, and helpful macOS utilities, Micah explores several approaches—from using advanced AI models to file automation—to minimize manual work for recipe digitization. Along the way, he shares personal automation workflows, recommends specialized apps, and offers time-saving tips for anyone facing similar tech challenges.

Key Discussion Points & Insights

1. Listener’s Challenge: Bulk Importing Recipes (00:00–04:05)

• Bob’s Question: Bob describes his labor-intensive system for converting Blue Apron recipe cards into Mela’s JSON file format—including scanning, OCR, prompting Copilot, copying JSON, and manually importing to the app.
• Bob’s Wish: “I would love to reduce the manual steps... I want to upload the entire set of PDF files and let the machines work for me.” (Bob’s letter, 01:31)

2. Enter Projects in AI Tools (04:06–07:40)

• Projects in LLMs: Micah introduces the concept of using custom "projects" in AI tools (Claude, OpenAI, Google’s Gemini Gems), allowing persistent instructions and batch processing.
  • “You are creating a sort of customized version of the LLM... it is influenced directly by a set of instructions that you give.” (Micah, 04:30)
• Personal Example: Micah shares his use case reformatting cluttered crochet patterns into his preferred style using Claude Projects, sparing him repeated manual clean up.

3. Step-by-Step Solution: Scaling Recipe Imports (07:41–11:32)

• AI Workspace Setup:
  1. Create a dedicated "Recipe Conversion" project in Claude (or other AI).
  2. Upload the Mela JSON spec so the AI learns the exact structure.
  3. Upload multiple PDF recipes in one go.
  4. Use persistent instructions: “Extract all recipe information and output valid JSON following the Mela recipe format, output each recipe as a separate code block…” (Micah, paraphrased, 09:53)
• Advantages: Removes repetitive instructions, enables batch processing, minimizes manual formatting steps.

4. Advanced Automation: End-to-End Workflow (11:33–14:35)

• Further Automation via Scripting:
  • Suggests using Hazel for macOS— an automation tool for file management:
    • “Hazel is...aware of your file system… you set up special rules and actions that the tool does based on its awareness...” (Micah, 12:03)
    • Example: Hazel watches a folder for newly scanned PDFs, triggers scripts to send files to the Claude or OpenAI API, collects the output JSON files, and organizes them into a designated import folder.
  • “All you would have to do then is take those files from that spot and bring them into Mela.” (Micah, 13:42)
• Cloud Integration: Store converted recipe files in an iCloud-synced folder for seamless import on iPhone.

5. App Alternatives for Non-coders (14:36–15:54)

• Specialized Recipe Apps:
  • Cookbook (for scanning and OCR)
  • Paprika and Recipe Keeper (“Paprika…can also import and scan,” 15:14)
  • Caveat: Must check export compatibility with Mela; may require interim conversion step.

6. Practical Tips for Reliable Automation (15:55–17:41)

• Leverage OCR: Since the office scanner already applies OCR, text extraction for recipe parsing is easier. But always:
  • “Check everything for errors… make sure you are scanning at a high resolution... so the OCR is able to do its best job.” (Micah, 16:31)
• Iterative Improvement: If frequent, invest time upfront in automation for ongoing time savings.

7. Broader Reflections on Automation’s Benefits (17:42–18:50)

• Micah emphasizes how today’s technology finally enables ordinary users to invent their own timesaving solutions.
  • “There are people who have a repeatable thing that they do… and goodness gracious, has it not become much more possible to come up with solutions that will actually work? And that’s what I love about this stuff.” (Micah, 18:19)

Memorable Quotes

• On Personalized Automation:
  • “Create a dedicated project that you can call Recipe Conversion... upload your PDFs in batches, because most of these AI systems are going to be able to process multiple files in a single conversation.” (Micah, 09:11)
• On Hazel’s Power:
  • “Hazel... can do that with so much. It can say when a file is added to this directory, when a specific kind of file is added to this directory...” (Micah, 12:37)
• On Setting Expectations:
  • “Now, of course, this requires some upfront setup, but the fact is that once it’s running, all you have to do is pop that stack of recipes into your scanner... and then just wait until those mailerecipe files are there ready for you.” (Micah, 14:06)
• On Tech Empowerment:
  • “Goodness gracious, has it not become much more possible to come up with solutions that will actually work? And that’s what I love about this stuff.” (Micah, 18:19)

Timestamps for Important Segments

• 00:00–04:05 Listener Bob’s workflow and pain points
• 04:06–07:40 Concepts of “projects” in Claude/OpenAI for repeatable automation
• 07:41–11:32 Specific setup steps using AI to export recipes in bulk
• 11:33–14:35 Deep-dive: macOS automation with Hazel and Python scripting
• 14:36–15:54 Recipe scanning app alternatives (Cookbook, Paprika, Recipe Keeper)
• 15:55–17:41 Scanning tips, OCR settings, error checking advice
• 17:42–18:50 Reflection on democratizing automation and tech empowerment

Takeaway

If you need to digitize a stack of recipes (or any structured documents), modern AI tools—configured as persistent “projects”—plus automation utilities like Hazel, can turn a tedious, manual process into a streamlined, nearly hands-off workflow. Invest a bit of setup time, and let the machines do the repetitive labor for you!

For more tech tips or to submit your own question, reach Micah at hot@twit.tv.