- AI Fire
- Posts
- 🛒 eBay Bans AI Bots
In partnership with
Meta paused its AI characters for teens. eBay drew a hard line against agentic bots. Claude crashed and took entire teams down for hours. And a new research paper just exposed how today’s best models actually “think” - and it might change how you prompt forever.
What's on FIRE 🔥
IN PARTNERSHIP WITH DEEL
Hiring in 8 countries shouldn't require 8 different processes
This guide from Deel breaks down how to build one global hiring system. You’ll learn about assessment frameworks that scale, how to do headcount planning across regions, and even intake processes that work everywhere. As HR pros know, hiring in one country is hard enough. So let this free global hiring guide give you the tools you need to avoid global hiring headaches.
AI INSIGHTS
A new analysis of 41M+ papers shows a sharp split:
AI-using scientists publish 3× more, get 5× more citations, and reach leadership earlier.
But their work clusters around the same “safe,” data-rich topics.
Across major fields, AI-heavy research:
explores less ground
concentrates around popular problems
generates weaker follow-on studies
James Evans (University of Chicago), who led the study, says the issue isn’t the models - it’s the incentives. AI rewards speed, not originality.
Journals are also seeing a rise in low-quality, AI-generated papers, pushing science toward volume over discovery.
Why it matters: AI is accelerating careers, but narrowing exploration. The challenge now is using AI to open new scientific questions - not just speed up the familiar ones.
PRESENTED BY GLADLY
AI that actually handles customer service. Not just chat.
Most AI tools chat. Gladly actually resolves. Returns processed. Tickets routed. Orders tracked. FAQs answered. All while freeing up your team to focus on what matters most — building relationships. See the difference.
AI SOURCES FROM AI FIRE
1. 12 Underrated ChatGPT Tricks that feel like cheat codes for 2026. We’ll show you the hidden 2026 tricks that most users never touch
2. Everyone generates cinematic B-Roll on demand in minutes (No stock footage). Don’t need to scroll through generic stock libraries just to create custom B-roll in <5 minutes
3. The ONLY 8 secret Goldmines to realistically REPLACE your 9-5 job (High RPM). Here’re 8 niches that actually generate real income and views for you. High RPM, evergreen topics, simple workflows
FIRE RECAP: BIGGEST AI NEWS THIS WEEK
🟦 Google rolled out Personal Intelligence across Gemini + AI Mode, letting it read Gmail, Photos, and history (opt-in) to answer personal questions and even inject Direct Offers into shopping flows.
📘 Anthropic dropped a new 57-page Claude Constitution, pushing deeper ethics rules, bans on harmful actions, and even hinting at moral status for advanced AIs - sparking a fresh debate.
💻 Microsoft is rolling out Claude Code company-wide, not just for engineers. PMs, designers, and more are now using it alongside Copilot as Azure access expands.
⚠️ A major Claude outage on Jan 22 crashed workflows across coding, support, and research - a reminder of how dependent teams already are on Anthropic day-to-day.
📰 The Guardian published a deep dive on Sam Altman and OpenAI’s “make-or-break” year, tying together the $1.4T infra push, regulation battles, Worldcoin, and whether OpenAI can actually make money at this scale.
TODAY IN AI
AI HIGHLIGHTS
🍎 Meta paused all teen access to its AI characters globally while it rebuilds a safer version ahead of multiple safety trials. Parents asked for tighter controls - Meta hit the brakes.
🛒 eBay banned all AI agents from shopping starting Feb 20. Bots can’t place orders unless eBay gives permission. A clear line drawn before agentic shopping goes mainstream.
🎥 Runway’s new Gen-4.5 fooled 90%+ of people in a real vs. fake video test. Nature clips were the hardest - most viewers guessed wrong.
🔍 Google expanded Personal Intelligence in Search so AI Mode can use Gmail + Photos to tailor recommendations. Only for AI Pro / Ultra, fully opt-in.
🎤 Google hired Hume’s CEO to boost emotional voice tech in Gemini, pulling in 7 engineers too. Voice is becoming the next big AI battleground.
💰 Big AI Deals: Leidos is teaming up with OpenAI to bring generative and agentic AI into U.S. federal workflows. The focus: real deployments in national security, health, and defense using OpenAI’s most secure models. Leidos, a $16.7B company, will build agentic workflows for threat analysis and deepfake detection.
NEW EMPOWERED AI TOOLS
👩💻 Cowork turns Claude into a digital coworker, assigning real tasks to files instead of endless chat.
🧠 Humans in the Loop is a free community for agentic coding, sharing insights on Claude Code, Cursor & MCP.
📚 BrainLoom converts PDFs into flashcards, unifying study with a local-first Learning OS.
🎙️ GetThis generates tasks from voice, text, or screenshots, organizing your brain dump instantly.
AI BREAKTHROUGH
A new paper from Google, UChicago, and the Santa Fe Institute says models like DeepSeek-R1, QwQ-32B, and OpenAI’s o-series don’t reason alone. They simulate multiple internal voices.
1️⃣ More conversation patterns
More Q&A
More perspective shifts
More internal conflict
This happens even with the same reasoning length.
2️⃣ Multiple hidden personas
Each trace shows 2–8 voices with different skills and traits.
3️⃣ One feature doubled accuracy
Boosting the “Oh!” realization feature pushed performance from 27.1% → 54.8%.
Takeaway: Top reasoning models think socially. If you prompt debate and perspective shifts, accuracy increases.
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply