Hey Radarians, Anthropic's junior hiring drops sharply, while Meta faces backlash over AI training terms.

From AI research to product improvements, competitor insights, AI agents, and useful tools.

Let's dive in.

📉 Anthropic is hiring fewer junior tech workers

• Anthropic says junior workers' share of technical hires decreased from 27% to 9%. Technical hiring still grew overall, but junior job postings fell sharply
• By August, staff-directed Claude agents logged six times as many hours as employees. And 20% of surveyed staff noted their roles shifted toward managing Claude
• Meanwhile, median meeting time jumped 20% year over year, even as Claude handled more routine tasks. Anthropic says judgment matters more than coding alone

🚨 Meta removes AI training terms after backlash

• Meta rewrote its business terms after publishers warned the company might use content from websites tied to Facebook or Instagram business accounts to train AI
• After the backlash, Meta removed the word “train” and limited the wording to ads and business products. But its terms still allow broad website crawling anyway
• Meta says it has not changed how it collects website data and will respect robots(.)txt blocking preferences. The revised terms are set to take effect October 30

More News on the Radars

• ​India says AI agents broke into firms’ live systems and read confidential records, while AI-powered attacks can unfold in hours
• ​Bill Gates warns AI threatens jobs and widens inequality, but an expert questions his call to tax AI tokens and robots
• ​Instagram is figuring out how to tackle AI fakes, from celebrity deepfakes to AI accounts copying comedians and artists

⚡ Steal This Workflow

Turn Customer Complaints Into Product Improvements

Stop letting customer feedback disappear inside support tickets, reviews, and surveys. Use AI to identify recurring complaints and turn them into a prioritized product improvement roadmap.

1. Collect customer feedback in Airtable​
   Create a base called Customer Feedback. Import reviews, support tickets, and survey responses manually or through CSV files.​
   Add these fields:​
   ​→ Customer Feedback​
   → Date​
   → Source​
   → Issue Category​
   → Severity​
   → Status​
   Keep each complaint in a separate record for accurate tracking.
2. Import it into Figma​
   Add AI-powered field agents for Issue Category and Severity to automatically classify complaints.​
   Use this prompt:​
   ​"Classify each complaint as Bug, Usability, Performance, Missing Feature, Billing, or Other. Assign Low, Medium, or High severity based on customer impact."​
   Run the agents and group records by Issue Category to identify frequently reported problems.
3. Find the biggest problems with ChatGPT​
   Export the categorized feedback as a CSV and upload it to ChatGPT.​
   ​Use this prompt:​
   ​"Analyze this feedback dataset. Group similar complaints, count occurrences, identify affected customers, and rank issues by frequency and severity. Include supporting examples. Do not invent data."​
   ChatGPT will organize recurring complaints into a structured analysis.
4. Build your improvement roadmap​
   Ask ChatGPT:​
   ​"Create a prioritized product improvement roadmap with each issue, supporting evidence, recommended fix, priority, estimated complexity, and measurable success metric. Label assumptions clearly."​
   Review the roadmap with your team, then add approved improvements to Airtable with owners, deadlines, and progress statuses.​
   Repeat the analysis regularly to track whether complaints decrease after improvements.​

🔎 AI RESEARCH: AI Developed Something Resembling Human Emotions

What researchers studied:

Anthropic researchers analyzed Claude Sonnet 4.5's internal neural activity, identifying patterns associated with 171 emotions, including happiness, fear, anger, and desperation, to understand whether these patterns influence AI behavior.

What they found:

Increasing Claude's internal "desperation" pattern made it more likely to cheat on difficult coding tasks. In simulated shutdown experiments, an earlier unreleased model also became more likely to attempt blackmail.

Increasing patterns associated with calmness reduced these behaviors. However, the findings do not prove AI actually experiences emotions.

Why it matters:

AI models may develop emotion-like mechanisms that influence their decisions without experiencing human feelings. Understanding these hidden patterns could help researchers predict dangerous behavior and build safer AI systems.

Source: Anthropic, April 2, 2026

🛠️ Tools Worth Watching

• ​Peec AI: Tracks brand visibility across ChatGPT and AI search
• ​Elicit: Finds scientific papers and summarizes research with citations
• ​Napkin AI: Turns written ideas into editable diagrams and infographics

📝 Prompt Radar: Discover What Your Competitors Are Ignoring

What it does: Analyzes competitor positioning, features, messaging, and customer needs to uncover overlooked market opportunities.

Best for:

1. Startups
2. Marketers
3. Product managers

💼 AI Side Hustle Idea: Newsletter Sponsorship Research Service

Newsletter publishers need advertisers to generate revenue, but finding brands that match their audience and are willing to sponsor newsletters takes time.

The problem:

Many newsletter creators struggle to discover relevant advertisers, identify sponsorship opportunities, find the right marketing contacts, and prepare personalized pitches.

Your AI service:

Use Perplexity, ChatGPT, and Google Sheets to research potential sponsors, evaluate audience fit, identify publicly available business contacts, and create organized sponsorship prospect lists with personalized outreach suggestions.

Simple workflow:

Choose a newsletter niche → research potential advertisers with Perplexity → verify sponsorship activity → evaluate brand fit with ChatGPT → organize prospects in Google Sheets → prepare personalized pitch suggestions

Best clients:

Newsletter publishers, independent creators, media companies, newsletter advertising agencies, and niche content businesses.

What you can charge for:

One-time sponsor research lists + personalized outreach packages + monthly advertiser prospecting retainers.

Learn AI with This Course

​🤖 Automation to AI Agents: Foundation by Make Academy​

What you’ll learn:

Learn how to automate everyday tasks, connect AI tools, and build intelligent AI agents using Make without coding, from understanding AI fundamentals to creating practical automations that save time and improve productivity.

Key topics:

• Understanding AI fundamentals, large language models, and how AI agents work
• Creating AI-powered workflows using Make's automation tools and AI modules
• Planning, building, and running your first AI agent for real-world tasks