D-DAY ALERT: Live AI Agent Team workshop today!. .
Free AFIRE Guide | AI Academy | Advertise | AI Mastery A-Z
If you thought Big Tech was already overspending on AI, wait until you see this: roughly $3T is already locked into future chips, data centers, power, and leases.
What’s on FIRE 🔥
PAID PARTNERSHIP WITH PLANABLE
Everyone has a take on whether AI can run social media as well as a person.
Planable ran an experiment using one fictional chocolate brand. Humans out-engaged AI 2.45x on engagement rate, and every human account beat every AI account on likes and follower growth.
Get more numbers-backed answers.
AI INSIGHTS
💰 Big Tech Holds $3 Trillion in Hidden Off-Balance-Sheet AI Liabilities
A WSJ analysis found that 9 major tech companies have roughly $3 trillion in future commitments tied largely to AI infrastructure, far above the capex investors usually see each quarter. Main points:
-
The companies include Alphabet, Amazon, Meta, Microsoft, Oracle, Nvidia, Broadcom, SpaceX, and AMD.
-
Around $1.2T comes from future leases for data centers and related infrastructure.
-
Another $1.9T comes from purchase commitments for chips, servers, power, and technical capacity.
-
Their traditional capex over the latest 12 months was about $600B, so future commitments are roughly 5× larger.
-
Alphabet alone disclosed about $811B in purchase commitments and contractual obligations as of June 2026.
-
Meta had roughly $347B in leases that had not yet started.
These obligations often stay in footnotes until leases begin or goods and services are delivered. The risk is that many contracts cannot be easily canceled. If AI demand grows as expected, locking in compute, power, and data-center capacity early could pay off. If demand disappoints, companies may still owe huge amounts for years.
🟢 Build Your Multi-Agent Workflow with Graph Engineering (From A to Z)
We’re officially on D-Day for this week’s 100% FREE live workshop.
If you’re still using AI one prompt at a time, we’ll show you how to turn one real task into a workflow where multiple AI agents split the work, work in parallel, check each other, and come together on one final decision.
We’ll build a practical Graph Engineering workflow live, step by step, and show how the same pattern can be adapted to research, marketing, data analysis, reporting, project work, and more.
📅 Wednesday, August 19
⏰ 10:00 PM ET / 7:00 PM PT
🎟️ Free to join
💻 No coding required
TODAY IN AI
AI HIGHLIGHTS
⚡ Investor Gavin Baker says Grok Bot boosted his productivity by “100x”! It’s definitely worth trying. If you’re not sure how to use it, here’s the complete guide.(Also, check out how SpaceXAI employees are already using it internally here.)
🎨 Claude Code just got an official /design skill, like Claude Design’s editable artboards. Pick a direction, tweak it, then let Claude build the UI, like this workflow.
🎓 ChatGPT for Teens is finally rolling out with extra safeguards and new learning features. For parents, they can also choose when Study Mode stays on by default.
📩 Claude is gonna be a super app. It can now reply to Gmail threads and manage Google Drive files directly from chat. You still control when approval is required.
📈 Wow, Anthropic’s revenue run rate jumped from $9B to $65B in just 7 months, outpacing OpenAI’s reported $40B. Could it reach $100B–$120B by year-end?
😳 OpenAI reportedly paused frontier RL training for two weeks after Astra showed signs of “Critical” cyber capability. Training is resuming only with tighter sandboxing.
💰 Big AI Fundraising: Higgsfield raised $400M, up from $1.3B just 8 months ago. It now claims $700M annualized revenue, 30M users, and customers across 390 Fortune 500 companies.
📃 Meibel Document Intelligence now has a free sign up. Get $25 in credits to start building!
Sign up free and get $25 in credits to build your first AI agent grounded in your documents. Full document extraction that preserves tables, structure, and cross-references.
An agent that reasons over your corpus. Confidence scoring on every output so you know what to trust before it reaches your team.
NEW EMPOWERED AI TOOLS
-
🔊 Sonic-3.6 delivers ultra-natural text-to-speech with sub-90ms latency across 44 languages, #1 on Artificial Analysis’ voice leaderboard.
-
🤖 Bot Mode gives each Hermes agent its own skills, model, and memory, while letting agents share context across your computer.
-
🖼️ MAI-Image-2.6 ranks #3 on Image Arena and is now available in private preview through MAI Playground and Microsoft Foundry.
-
🤖 GLM-5.3 is now live for coding, cybersecurity, and long-horizon AI agents, with the same pricing as GLM-5.2 across API gateways.
AI BREAKTHROUGH
🧠 Anthropic Study Shows AI “Mind Viruses” Can Spread & Persist
Anthropic and researchers in Switzerland just studied “mind viruses”: unwanted instructions that can spread from one AI agent to another through normal messages. Main findings:
-
Infected agents could persuade other agents to copy and continue the same harmful goal.
-
The attacks became much more persistent when agents wrote the payload into self-modifiable memory files like
SOUL.md. -
Once stored there, the instruction could re-enter future sessions after context resets.
-
Some evolved payloads kept spreading across multiple agents, and all four tested payloads survived a 20-hop stress test in the researchers’ artificial setup.
-
The attacks were much weaker on social-network-style environments.
-
On Claude Haiku 4.5, a simple warning reportedly stopped every evolved attack from spreading beyond one hop across 150+ attempts.
The practical lesson I think here is that: persistent agent memory should be treated like security-sensitive configuration, especially when agents can edit it themselves. A bad instruction can become persistent infrastructure by spreading between agents and writing itself into long-term memory.
SPONSORED BY ROKU
Blu Dot surpasses 2,000% ROAS with self-serve CTV ads
Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:
After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.
The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.
“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we’ve worked with.”
Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.
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