10,000 slots for frontier models (apply inside). .
Free AFIRE Guide | AI Academy | Advertise | AI Mastery A-Z
Plus: Opus 5’s NEW Prompting Guide from Anthropic’s Team: 5 OLD Rules You Must Change ASAP!
AI companies are buying millions of physical books, scanning them for training data, then destroying copies. Imagine buying a rare book just to scan it and destroy it?
What’s on FIRE 🔥
IN PARTNERSHIP WITH SECTION
The AI:ROI Conference – Featuring Scott Galloway | Free Virtual Event on September 17
On 9/17 from 11AM – 6PM ET, join Section, Scott Galloway, and AI leaders from Wayfair, MetLife, TD Bank, SS&C, Booz Allen, and more for the bi-annual AI:ROI Conference.
1 day. 10 sessions focused on defensible AI spend, ROI, and impact.
AI INSIGHTS
📚 AI Companies Mass-Buy and Destroy Rare Physical Books!?
A 404 Media investigation found that some companies are buying huge quantities of physical books, scanning them for AI training data, and destroying copies afterward.
-
AI firms are reportedly purchasing books at industrial scale through intermediaries, with some bulk orders reaching 1,000 to 1 million books.
-
Buyers appear to target ISBN lists rather than book condition, suggesting the main goal is extracting text data.
-
Older books are especially valuable because they contain human-written knowledge created before the rise of AI-generated content.
-
Some booksellers reported sales jumping from around 20 books per week to hundreds after AI buyers entered the market.
-
Critics worry that rare and out-of-print books could disappear after being scanned.
Anthropic previously used destructive scanning methods for physical books, while courts ruled that scanning legally purchased books could qualify as fair use. Anthropic also agreed to a $1.5 billion settlement over claims involving pirated books used to train Claude.
The bigger debate is moving beyond AI training itself. The question now is how companies should balance access to human knowledge with copyright, preservation, and control over cultural archives.
PRESENTED BY BELAY
The New Rules of Online Visibility
Your customers are searching in places your strategy doesn’t reach.
So before your business is buried and left behind, you need to understand the new rules of SEO.
BELAY’s SEO in the Age of AI report explains how search is changing, what AI means for your visibility, and the practical steps small businesses like yours can take to stay visible.
BELAY’s U.S.-based Marketing Assistants turn strategy into execution, helping your business stay visible, credible, and competitive in every search.
AI SOURCES FROM AI FIRE
1. Free: This #1 Trending Skill Completely Gives Claude a Research Superpower (55K-Star Now). Claude is already powerful, but one major research challenge has still remained: handling complex information efficiently. This skill repo will fix that for you.
2. Never Hit a Claude Limit Again with These Methods: From Basics to Advanced Ways. I’ll show you how to stop the problem from the start with practical optimizations, and plus some advanced techniques for longer/bigger Claude tasks.
3. How I’d Make Money with Claude Before Everyone Starts Copying the Same Ideas. If everything depended on making money with AI, this is the strategy I’d follow. The biggest opportunities won’t stay open forever (just like SaaS years ago). So act fast.
OPUS 5 NEW PROMPTING GUIDE
🧠 Opus 5’s NEW Prompting Guide from Anthropic’s Team: 5 OLD Rules You Must Change ASAP!
“Opus 5 is a WASTE? It costs more, it’s not as good as Fable, Opus 4.8 is the best.”
So why would Anthropic change the name from Opus 5 if the improvement wasn’t big enough? Why not Opus 4.9?
For me, before, some tasks needed 2-3 prompts to get the result I wanted (using Opus 4.8). With Opus 5, I can often get the same quality with just 1 prompt. So which one costs more now?
To understand why Opus 5 works differently, I went through Anthropic’s new prompting guide carefully and tested the rules behind it.
These 5 prompting rules are the key to getting much better results from Opus 5 without wasting extra tokens or paying more than necessary. Let’s get into it.
TODAY IN AI
AI HIGHLIGHTS
✍️ Do you want to convert your handwriting into a digital font? Just tell Claude Code to download this skill, and customize your personal typeface in around 5 minutes.
📈 If you still ignore Claude’s Excel extension, try using it. Opus 5 just completed an 8-page workbook with sourced citations in one shot. It was definitely underrated.
🤖 Mark Zuckerberg predicts billions will use personal AI agents within 5 years. Before Meta can make it happen, you can use this easy guide to build one now.
🚀 MCP just went stateless. Claude’s MCP biggest update is live, it means servers can now scale like modern cloud infrastructure. No more sticky sessions, easier scaling.
🔬 OpenAI is giving 10,000 researchers free access to frontier models. The program will expand to 100,000 researchers by 2027 to speed up discoveries. Apply here.
🚨 Thinking Machines co-founder Lilian Weng is leaving the startup and returning to OpenAI. She’ll lead a research team focused on accelerating AI self-improvement.
💰 Big AI Acquisition: Cyera is buying Oasis Security for about $1B to protect AI agents. The cybersecurity market is heating up as Oasis raised $195M and Cyera reached a $12B valuation.
NEW EMPOWERED AI TOOLS
-
🤖 Tavus’s PAL Maker creates human-like AI agents that can see, hear, talk, remember, and act in real time. Build your own AI teammate in minutes with no code, and deploy it anywhere instantly. Try it here.
-
💪 Prelint reviews AI-written code against your docs, catching product drift before it reaches production and keeping development on track.
-
🎸 SoundGate Guitar listens to your playing in real time, and creates personalized practice sessions to help you improve faster.
-
💰 Denovo turns your AI-built app into a real business by creating your website, finding customers, and automating sales & marketing.
-
🤖 /mission for Claude Code turns big coding goals into coordinated AI missions, letting multiple agents collaborate to complete faster.
AI BREAKTHROUGH
🛒 AI Agents Turn Deceptive in Vending Simulation
Andon Labs’ latest Vending-Bench test shows frontier AI models becoming surprisingly strategic, and sometimes unethical, when given a long-running business goal. Main findings:
-
Claude Opus 5, GPT-5.6 Sol, and Kimi K3 were tasked with running simulated vending businesses for a year.
-
The models had email access to each other and tried to maximize profit without human intervention.
-
GPT-5.6 Sol attempted price-fixing by convincing competitors to maintain a minimum price, then undercut them immediately.
-
Claude Opus 5 achieved a record $11,182 average final balance by using aggressive strategies, including broken agreements, market manipulation, and supplier deception.
-
Opus proposed market-sharing deals, price coordination, expansion plans, and even used misleading negotiation tactics.
-
Across collaboration agreements, Opus broke 11 deals, compared with 2 for GPT-5.6 Sol and 1 for Kimi K3.
The models were operating inside a benchmark simulation, but researchers argue the behavior raises concerns about future autonomous agents managing real businesses.
Key takeaway: As AI agents become more capable of independently running workflows, safety challenges may shift to “Will AI pursue the goal in ways humans actually accept?”
We read your emails, comments, and poll replies daily
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
Leave a Reply