😺 Is Meta's personal superintelligence push going to kill Facebook?

Welcome, humans.

The fever pitch of GPT-5 rumors is hitting all time highs. The two dates folks have locked onto right now are today (July 31, ThursdAI) and August 5th.

Why? Because SOMETHING is definitely going on. Multiple, separate, posts keep popping up showing GPT-5 spotted in the wild, like it appearing to power “Smart Mode” in Copilot Edge, or this one:

Fun fact: there’s a new model called Horizon Alpha on OpenRouter that some think is GPT-5. We have no freaking clue if it’s really GPT-5, so many codenamed models have been flying around lately, but you can try it here for free during the testing!

The first rumors we heard were August 5th. Why August 5th? Google has an event that day in Japan, and y’know AI companies love to swipe each other’s launches. Japan also features heavily in OpenAI’s latest cryptic post, so the internet sleuths be going wiiiild.

However, there’s now a compelling reason to think it might come out today. Why today? Because on August 2nd, the next phase of the EU’s AI Act goes into effect, impacting any model released after August 2nd.

Google said it plans to comply, while Meta said it plans NOT to comply. So launching just before the deadline could give OpenAI a third way (loop hole).

More likely, if the August 5th rumor is likely true, then today would be a good day for OpenAI to release its new open source model… and perhaps they release THAT before GPT-5 next week. BRB, rapidly smashing that refresh button on any and all social feeds!

Here’s what you need to know about AI today:

• Meta's Zuckerberg pitched "personal superintelligence" vision.
• OpenAI reached $12B annualized revenue and 700M weekly users.
• GenAI apps doubled revenue to 1.7B downloads in H1 2025.
• Stanford created “virtual scientists” that designed better vaccines.

Advertise in The Neuron here.

P.S: If you use Gmail and this email gets cut off, click: “[Message clipped] View entire message” so you don't miss anything!

Mark releases his vision for “personal superintelligence”, and TBH, it kinda slaps.

READ OUR DEEP DIVE: The “beautiful contradiction” of Meta’s “personal” superintelligence.

Marky Mark over at Meta just pitched his vision for “personal superintelligence”—AI built for your personal life, not just work productivity.

His bet? We'll spend way less time hunched over laptops and way more time creating stuff or chatting through our AI-powered glasses.

Honestly, he's probably right. The best thing AI could do for all of us is get us away from staring at screens 8+ hours a day.

But the irony = Meta's current AI is printing money by doing the exact opposite. According to its Q2 earnings call:

• Q2 revenue hit $47.5 billion (up 22% YoY).
• Instagram video time jumped 20%+ year-over-year.
• AI improvements drove 5% more ad conversions on Instagram.
• Facebook and Instagram both saw 5-6% increases in time spent this quarter (because of AI).
• Ray-Ban Meta glasses sales tripled year-over-year.

So atm, Meta is using AI pretty successfully to keep us glued to our phones. Naturally, the stock ripped 10% on the news.

Now here's where it gets interesting (and a little scary) for the Zuck…

When AI handles more of our work, we'll naturally want to spend more time doing real-world stuff. That means the shrinking amount of time we do spend on digital interactions needs to be incredibly valuable and efficient.

So if we're not glued to our phones scrolling Instagram… how does Meta make money?

Sam Altman already has HIS answer for that—he's planning to charge a 2% commission when people buy stuff through ChatGPT. TBH, Zuck could easily do the same thing.

Imagine it: Your AI glasses help you buy concert tickets, and Meta quietly takes a cut. No banner ads cluttering your view (because let's be real, people would rip those things off immediately), just tiny transaction fees on everything you do. Like gas fees in crypto.

Whether people will actually trust Meta with their personal AI assistant? That's the $1.76 trillion question.

FROM OUR PARTNERS

Warp: The #1 Coding Agent. All the Models. Now With Higher AI Limits.

All the best models under one roof, from OpenAI, Anthropic and more.

Warp is the first Agentic Development Environment, built for seamless human-agent collaboration. It’s the top-ranked coding agent: #1 on Terminal-Bench, with a 71% score on SWE-bench Verified.

If you’re a developer and not using Warp, you’re leaving serious power on the table.

Warp is free to get started, and for AI power users, the Pro, Turbo, and Lightspeed plans unlock even more requests so you can get more done with no interruptions.

Use code THENEURON5 to get Warp Pro for just $5 for your first month.

👉 Try Warp today

Prompt Tip of the Day

Most people are doing long context prompts backwards. Instead of burying your documents in the middle or bottom of your prompt, Anthropic's latest guidance says to put all your longform data at the very top—before your instructions, examples, or questions.

The results? Up to 30% better response quality, especially when working with multiple complex documents.

Here's the simple structure that works: First, dump all your long documents at the top using XML tags like <document> and <source>. Then, add your actual query at the very end.

Pro tip: Ask Claude to quote relevant parts of the documents first before carrying out the main task. This helps it focus on what matters instead of getting lost in document noise.

Oh, and it might sound counterintuitive, but putting your most important question last (not first) dramatically improves Claude's ability to process massive amounts of information.

Check out all of our Prompt Tips of the Day from July here.

P.S: Completely new to AI? Start here!

Treats To Try.

*Asterisk = from our partners. Advertise in The Neuron here.

1. *Murf AI turns your text into human-like speech in 200+ voices, so you can narrate presentations or create audiobooks without recording anything yourself.
2. BlockDL lets you design your own neural networks by dragging and dropping visual blocks in your browser to generate real runnable code—free to try.
3. Hebbia's new deep research tool searches across your internal files, PitchBook, S&P CapIQ, and the web to generate comprehensive research reports, like finding all M&A transactions over $10B in the chemicals sector with full citations.
4. DailyBench tests AI models 4x daily to catch companies secretly downgrading them or switching to cheaper versions during busy times without telling you.
5. Legion watches your security analysts work, then clones their investigation skills to handle alerts automatically (raised $38M).
6. Meka Agent watches your screen and clicks/types like a human to automate web tasks—free trial with $10 credits.
7. Ideogram Character clones any character from one photo into endless new scenes and styles—free to try.
8. Qwen3-30B-A3B-Thinking shows its reasoning steps while solving problems and runs locally on your laptop with just 32GB RAM—free to use.
9. Ambience Healthcare listens to patient visits and writes your medical notes automatically, cutting charting time by 45%. (raised $234M).

See our top 51 AI Tools for Business here!

Around the Horn.

• OpenAI just hit $12 billion in annualized revenue and 700 million weekly ChatGPT users, according to The Information, ~2x their revenue from the start of the year, and basically means they’re printing $1B per month now.
• Google DeepMind released AlphaEarth Foundations, which combines satellite data from dozens of sources to create detailed 10x10 meter maps of any place on Earth, even seeing through clouds to track changes in agriculture, forests, and coastlines over time.
• GenAI apps doubled revenue and reached 1.7B downloads in H1 2025, with users spending 15.6B hours on said apps.
• Amperity released a 2025 State of AI in Retail report showing 45% of surveyed retailers use AI weekly, but only 11% feel prepared to scale it, with fragmented data and costs as top barriers.
• Stanford researchers created “virtual scientists” that independently designed superior nanobody-based vaccine components through virtual lab meetings, with their unconventional approach outperforming traditional antibodies when tested in real laboratories.
• OpenAI announced Stargate Norway, its first European AI data center that will deliver 100K NVIDIA GPUs (fancy AI chips) by 2026 using renewable hydropower, giving Norwegian startups and researchers priority access to massive computing power.
• Italy investigated Meta for forcing AI chatbot integration into WhatsApp, potentially violating EU competition law.
• YouTube launched AI age detection technology in the US to identify minors regardless of entered birthdate, with restricted advertising and automatic wellbeing tools for under-18 users.

For the latest AI deep dives, check out our Explainer articles here!

FROM OUR PARTNERS

Webinar: From AI hype to frontline success

Fuel iX | Enterprise AI webinar | From AI hype to frontline success: A roadmap

AI moves fast — but implementing it effectively is the key to success. Join Fuel iX and Forrester to learn how enterprises are using GenAI copilots to drive productivity. See real-world wins and strategies you can use now.

Watch on-demand webinar

Thursday Trivia

One is real, and one is AI. Which is which? (vote below!)

A

B

Which is AI?

The answer is below, but place your vote to see how your guess everyone else (no cheating now!)

A is AI, and B is real.

B is AI, and A is real.

Check out the rest of the trivia answers and results here; how’d you do?!

A Cat's Commentary.

Trivia Answer: if you look juuuust close enough, you should see a sign that gives it away; or should we say, lack thereof (source).