GPT-6 teaser: "People want memory"

👋 Good morning/evening (wherever you are). It’s Thursday.

Altman says GPT-6 will come faster than the gap between GPT-4 and GPT-5, with memory as its defining feature. The vision is for ChatGPT to remember who you are, adapt to your preferences, and become truly personal.

He acknowledged GPT-5’s rocky rollout and admitted memory still raises privacy risks since it isn’t encrypted. Looking ahead, Altman pointed to neural interfaces and new energy infrastructure as critical areas alongside model development.

(Source: CNBC)

OK let’s keep going ↓

🔥 Creative pulse

This AI-generated film is ridiculously good…

Lion El Aton is the best filmmaker you’ve never heard of.

I love his style because it’s bold and built on conviction. You can feel it in what he chooses to release.

Watch the video. Posted on LinkedIn.

PS: If you’re passionate (and obsessed) about AI creative…guess what?! I just launched an AI creative club. It’s only $5/month. We test new tools, share experiments, and learn together in a supportive, judgment-free space. Join here.

💡 Here’s what you should know

• Google expands AI Mode worldwide

  The feature now reaches 180 countries with new agentic tools like reservations and personalized results.

• GSA and Google launch ‘Gemini for Government’ deal

  The OneGov agreement gives U.S. federal agencies access to Google’s Gemini AI models, cloud services, and agentic tools at steep discounts through 2026.

• Nvidia halts China-focused H20 AI chip production

  The company told suppliers including Amkor and Samsung to suspend work on the H20, its AI chip designed for the Chinese market, according to The Information.

• DeepSeek releases V3.1 with hybrid inference and agent skills

  The update introduces Think and Non-Think modes, faster reasoning, stronger tool use, 128K context, and open-source weights, positioning V3.1 as a step toward the agent era.

• Perplexity loses bid to toss News Corp copyright suit

  A New York judge rejected Perplexity AI’s attempt to dismiss or transfer the case brought by Dow Jones and the New York Post over alleged misuse of articles for AI training.

• Anthropic forms education board, launches AI courses

  The company formed a Higher Education Advisory Board led by former Yale president Rick Levin and introduced three Creative Commons AI Fluency courses to help educators and students integrate AI responsibly into learning.

• Anthropic and U.S. DOE develop nuclear safeguards for AI

  In partnership with the National Nuclear Security Administration, Anthropic built a classifier that flags potentially harmful nuclear queries with 96% accuracy, aiming to prevent misuse of AI models for weapons development.

• Wired, Business Insider pull AI articles by fake freelancer

  At least six outlets removed stories written under the name “Margaux Blanchard” after discovering the articles were AI-generated, raising new concerns about newsroom verification.

• AI-enhanced Wizard of Oz will debut at Las Vegas Sphere

  Warner Bros and Google reimagined the 1939 classic with AI-powered visuals and immersive effects, set to premiere on Aug 28 inside the 22-story Sphere venue.

Update on Suno Studio: The company has been teasing its new DAW, and early previews show an interface in the style of GarageBand or Logic, but with AI built in.

💰 The numbers

• Anthropic, previously reported to be raising $5B at a $170B valuation, is now in talks to raise up to $10B, making it one of the largest megarounds ever for an AI startup.

• Meta signed a six-year, $10B+ deal to run its AI workloads on Google Cloud.

• Nuro raised $203M Series E at a $6B valuation with new backing from Nvidia and Uber.

• Wistron will invest $761M to build two AI supercomputer manufacturing plants in Fort Worth, Texas, creating over 800 jobs.

• Twin Health raised $53M Series E for its AI metabolic health platform.

• Wisdom raised $21M Series A to bring AI automation to dental revenue cycle management.

• SynergySuite raised $12M to scale its AI-powered back-of-house platform for restaurants.

• Definite raised $10M seed to build an AI-native data stack for enterprises.

• Leaping AI raised $4.7M seed to scale its AI voice agents, already handling 10,000 calls per day with 90%+ customer satisfaction.

🧠 Thought starters

How much energy does one AI prompt really use?

Google just shared the first per-prompt energy numbers for Gemini. The median text query: 0.24 watt hours (about the same as running a microwave for one second).

Key takeaways from the MIT article:

• 58% of energy is from AI chips, the rest comes from CPUs, backups, and cooling

• Each prompt also uses about 0.26 ml of water and emits 0.03 g CO₂

• Efficiency has improved 33x in the past year

• The figure reflects median text prompts, not heavier reasoning or image/video generation

• Google did not disclose total daily queries, so overall usage remains unclear

Something to think about:

Will this spark a standardized AI energy score, or remain selective reporting controlled by Big Tech?

😂 Meme of the day

Thanks for reading,

Eddie

Join the AI Creative Club (you’ll find me there). It’s not only for artists and filmmakers…it’s for anyone who recognizes this once-in-a-lifetime opportunity to learn as much as you can about AI to multiply your impact...to become superhuman. Remember, creativity is just intelligence having fun. This club is for ultra curious people, and if you’ve made it this far…you’re pretty damn curious and probably the smartest one in your circle of friends. It would be my honor to have you in there.

Are we friends on LinkedIn and X? Let’s connect. DM me or hit the follow button.

P.S. If this was valuable, forward it to a friend. If you’re that smart friend, subscribe here.

P.P.S. Interested in reaching other ambitious readers like you? To become a sponsor, reply to this email.