• The Artificial Newsletter
  • Posts
  • 🧠 Automating Intelligence: Build a ReAct-Style Agent That Writes and Shares AI News on LinkedIn

🧠 Automating Intelligence: Build a ReAct-Style Agent That Writes and Shares AI News on LinkedIn

— Issue #13 of The Artificial Newsletter

In this week’s edition, I want to take you behind the scenes of an automation I recently built using Zapier + OpenAI, which not only fetches trending AI stories but also summarizes and publishes them as a LinkedIn post — without me writing a single line of copy each day.

Let’s walk through the technical build and why this qualifies as a ReAct-style agent, even though it’s completely no-code.

⚙️ What It Does

Every few hours, the agent:

  • Fetches AI news from curated RSS feeds (e.g., TechCrunch, VentureBeat, MIT Tech Review)

  • Filters and formats the top 3 headlines using a small JavaScript snippet

  • Sends the context to GPT-4o to generate a compelling LinkedIn post

  • Emails me the draft (for review or backup)

  • Automatically publishes it to LinkedIn using Zapier's LinkedIn integration

🤖 Why It’s a ReAct Agent

The term ReAct (Reasoning + Acting) comes from a 2022 paper by Google DeepMind and OpenAI researchers, where agents:

  • Reason through language using LLMs

  • Act by invoking external tools or APIs

While Zapier doesn’t support true step-by-step LLM reasoning natively, this build simulates it well:

Step

Role

Tool Used

Behavior

1

Perception

RSS Parser

Pulls latest events (inputs)

2

Reasoning

GPT-4o

Summarizes and selects what matters

3

Action

Zapier + LinkedIn

Publishes autonomously

4

Reflection

Gmail

Archives/logs the step

Zapier Setup

Email step is optional**

Technical Highlights

✅ Step 1: RSS by Zapier — New Item in Feed

🔹 App: RSS by Zapier

🔹 Trigger Event: New Item in Feed

This trigger runs whenever a new article is published in a specified RSS feed.

🔧 Configuration Fields:

Field

Value

Feed URL

Use one or more of these curated AI-focused RSS feeds:

https://techcrunch.com/tag/artificial-intelligence/feed/   https://venturebeat.com/category/ai/feed/   https://www.technologyreview.com/feed/   https://www.analyticsvidhya.com/blog/category/artificial-intelligence/feed/ 

(Use one to start. You can use a tool like Feedly + Pipedream later for merging multiple feeds.)

| How often do you want this to check for updates? | Choose: Every 2 minutes, 15 minutes, or Hourly (based on your Zapier plan) |

✅ Step 2: Code by Zapier — Run JavaScript

🔹 App: Code by Zapier

🔹 Event: Run JavaScript

🔧 Input Data (Required)

Add 3 inputs, assuming you’re triggering this Zap when 3 RSS items are available (either manually structured or via Digest/Sheet).

Input Name

Value (Dynamic Field from RSS Feed)

title1

Title from 1st RSS item

summary1

Summary or Description from 1st RSS item

title2

Title from 2nd RSS item

summary2

Summary from 2nd RSS item

title3

Title from 3rd RSS item

summary3

Summary from 3rd RSS item

Use the ➕ icon in each field to select these from either individual feed steps or a Digest/GSheet setup.

🔧 JavaScript Code

Paste this into the code editor:

const articles = [   { title: inputData.title1, summary: inputData.summary1 },   { title: inputData.title2, summary: inputData.summary2 },   { title: inputData.title3, summary: inputData.summary3 } ];  const combined = articles.map((item, index) => {   return ${index + 1}. ${item.title} - ${item.summary}; }).join('\n\n');  return { combinedNews: combined }; 

✅ Output

  • Output variable name: combinedNews
    This will be passed to OpenAI in the next step.

🔁 If you're only getting one article per trigger and want 3 total, you'll need to use a Digest step before this, or temporarily mock the title2, title3 inputs with test data for OpenAI.

Great — let’s configure Step 3: ChatGPT (OpenAI) to generate the LinkedIn post from the combined news.

✅ Step 3: ChatGPT (OpenAI) — Conversation

🔹 App: ChatGPT (OpenAI)

🔹 Action Event: Conversation

🔧 Configuration Fields

User Message

Paste the following prompt exactly:

Based on the following top 3 AI news stories, write a single LinkedIn post that:
- Is engaging and informative (max 300 words)
- Highlights key trends or takeaways
- Adds 2–3 relevant hashtags
- Ends with a question or CTA to the audience

Here are the news items:

{{combinedNews}}

⚠️ Use the ➕ icon to insert combinedNews from Step 2: Code by Zapier

Model

Choose: gpt-4o (recommended) or gpt-3.5-turbo

(Optional) System Message

If this field is visible:

You are a professional LinkedIn content strategist.

(Optional) Temperature, Max Tokens, etc.

If available:

  • Temperature: 0.7

  • Max Tokens: 400

  • Top P: 1

  • Frequency Penalty: 0

  • Presence Penalty: 0.3

If not visible, Zapier handles these with default values.

🧪 Output Field to Use

In the next step (Email), you'll reference:

3. Conversation in ChatGPT (OpenAI)Reply

This contains the generated LinkedIn post.

Perfect — let's wrap it up with the final two steps: Email (to send yourself the draft) and LinkedIn (to optionally auto-post it).

✅ Step 4: Email (Gmail or Outlook) — Send Draft to Yourself

🔹 App: Gmail (or Email by Zapier, or Outlook)

🔹 Action Event: Send Email

🔧 Configuration

Field

Value

To

your email

Subject

LinkedIn Draft: AI News Summary

Body

Insert → 3. Conversation in ChatGPT → Reply

💡 Click the ➕ icon and search/select: Reply from the ChatGPT step — this is the LinkedIn post GPT generated.

🧪 Test This Step

You should receive an email with your LinkedIn post formatted and ready to review.

✅ Step 5: LinkedIn — Auto Post (Optional)

⚠️ This step only works if you have LinkedIn connection enabled in Zapier (requires LinkedIn Company Page or Creator account).

🔹 App: LinkedIn (Zapier Integration)

🔹 Action Event: Create Share Update or Create Company Update

🔧 Configuration

Field

Value

Organization

Your personal or company LinkedIn profile

Content

Insert → 3. Conversation in ChatGPT → Reply

Visibility

Public (default)

If it's a company page, choose “Create Company Update” instead and select your page name.

🎉 You're done! You now have:

  • ✅ An autonomous ReAct-style content agent

  • 📥 Daily email drafts of AI news posts

  • 📢 Optional 1-click or auto-posting to LinkedIn

🔄 What’s Next?

  • Add a GPT-based moderation layer to rate content quality

  • Archive each post in Notion or Google Sheets

  • Offer this as a service for founders and teams who want automated thought leadership

🧩 If you're building AI-driven automations or experimenting with ReAct-style agents — I'd love to hear from you. Reply to this or drop me a DM.