For several months, I’ve casually been researching and testing how to automatically generate article summaries based on links in an RSS feed in order to curate content for my Menopausey communities. I’ve got Google Alerts set up and have those going via RSS feed to a Slack channel. I also have the most recent ten items per topic posting to a WordPress page using the (free) WP RSS Aggregator. I kept getting stuck trying to be able to pull full content using other tools when limited to Google Alerts/its RSS.

Rewind for a couple of decades with me…in grad school, I had to take an undergrad level macroeconomics course even though it seemingly had nothing to do with my degree program. It was a giant lecture hall course and not my preferred mode of learning at all. I came out of it with an incredibly valuable lesson, though. TIME IS VALUABLE. Sometimes all we have is time (vs. money) and we don’t have a choice to invest in solutions that might buy lots of time back. And sometimes, it makes sense to spend that money b/c the time savings is too great.

That’s the position I found myself in today. I again tried to automate this process and was on the verge of giving up. It’s so frustrating when you know there’s a solution and you’ve exhausted all possibilities to figure it out. I explored various platforms’ documentation libraries, chatted with ChatGPT and Poe, Zapier’s own AI tools, watched YouTube videos, and ran various tests. One of the chats pointed me to Feedly’s Pro+ account and ability to export to RSS. While that didn’t turn out to solve my challenge, I did ultimately get all of this to work by integrating Zapier with Feedly, OpenAI, and Slack. 

(Note, this requires paid accounts and some slightly technical abilities, but can be accomplished using no-code tools and an API key.)

Startups can apply for generous credits through Microsoft for Startups, AWS Activate, and Google for Startups. They all come with great perks, but Microsoft for Startups also includes the ability to apply for OpenAI API credits and Azure’s OpenAI Service. (You can redeem these offers from the Microsoft for Startups Founders Hub portal once you’re accepted.)

I found that all three programs were responsive to my applications in a matter of days and I’ve accumulated more than $10K in credits for various tools at this point as a pre-product, pre-revenue, solopreneur, side startup. That’s pretty amazing and is greatly accelerating my workflow.

I haven’t come across startup credits for Zapier or Feedly, however, so I invested in annual accounts for both. Note that both have free tiers, however, I needed the paid plan functionality for this. $19.99/month for Zapier Professional and $12/month for Feedly Pro+ when paid annually for each. I use the free version of Slack for now.

Now On to the Solution…

  • Create new AI feed in Feedly
  • Create new Zap
    • Connect Feedly account
    • Connect OpenAI account
    • Connect Slack account

Here’s what the Zap looks like:

A flowchart demonstrating a sequence of actions in an automation process is shown. The steps are: 1) New Article in Feed in Feedly 2) Parse Webpage in Web Parser by Zapier 3) Summarize Text in OpenAI (GPT-4, DALL-E, Whisper) 4) Send Channel Message in Slack (Alt text generated by ChatGPT and reviewed by the human author of this post for accuracy)

For “New Article in Feed in Feedly,” I selected the new AI Feed I mentioned above as the Trigger.

For Parse Webpage in Web Parser by Zapier, I selected article URL to parse, and outputted to Plaintext.

I used gpt-3.5-turbo-instruct for the model and selected Title:, Content:, and Excerpt: for Source Text in the Action for Summarize Text in OpenAI. I left the other defaults alone.

I selected the following for Send Channel Message in Slack and updated formatting slightly to make the messages pop a bit more.

A screenshot of a form titled "Message Text" with the following fields:

"2. Title: Prepping for Me...6 Ways To Start"
"1. article source name: 'menopause' - Google News"
"3. Summary: Preparing for m...ore manageable."
"2. Url: https://www.wel...e-for-menopause"
(Alt text generated by ChatGPT and reviewed by the human author of this post.)

Here’s what things look like in Slack:

So What, Now What?

I can quickly scan headline summaries to better determine which articles I should spend time reading and which are good candidates to be shared in Menopausey communities, like the WhatsApp group I’m building up, LinkedIn, and Menopausey’s Farcaster/Warpcast channel.

Additional Resources

Header image generated by Adobe Firefly. Prompt: neon colors infinity maze where it’s easy to get lost

Categories: AIChat GPTDataOpenAI