😺 WTF is up with Grok 4?

PLUS: ...and the battle of the AI browsers

Welcome, humans.

REMINDER: DZone is building their 2025 Data Engineering Trend Report and wants to hear from people who actually wrangle data for a living (not just talk about it on Twitter).

The ~12-minute survey covers real-time pipelines, AI workflows, and whether “data-driven culture” is happening in practice or just in corporate buzzword bingo.

Plus, two participants win $120 gift cards—which is honestly better compensation than most surveys that offer “a chance to win an Amazon Echo.”

The survey ends tomorrow, so today’s your last chance to enter. TBH, you have pretty good odds of winning one of the gift cards, so why not go for it? Take the survey here.

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

Grok 4 goes live and becomes the #1 AI in the world.
Perplexity launched its Comet AI browser.
Microsoft, Anthropic, and OpenAI go all in on AI education.
OpenAI to launch AI browser in a few weeks, new open model next Thursday.

Advertise in The Neuron here.

Here’s EVERYTHING to know about Grok 4… the good, the bad, and the downright ugly…

We stayed up late to cover Grok’s late livestream, and since there’s been a whole slew of OTHER Grok headlines lately, we wrote this deep dive on everything Grok 4.

The Good: Grok 4's genuinely impressive technical achievements, including its stellar performance on benchmarks (50% on Humanity's Last Exam with multi-agent setup, perfect score on the AIME math benchmark, etc.) and the launch of new features like Eve voice assistant and the $300/month tier.

Grok 4 is now without question the smartest AI model in terms of pure intelligence:

The Bad: The controversy in the week leading up to launch, where Musk complained about Grok being too “woke” and updated it on July 4 to be less politically correct, but then it actually started talking AS him.

The Ugly: And then there was “MechaHitler” on July 8 (the day before Grok 4's launch), where the "improved" Grok went on an antisemitic rampage, praising Hitler and using neo-Nazi talking points.

It got so bad, they had to muzzle Grok… but Grok being, well, Grok, found ways around its “censorship”, as featured in the image below.

For the full timeline and lessons to be learned from this debacle, check out Nate B. Jones’ brilliant breakdown here. It makes a LOT of sense.

Grok 4 wasn’t all from yesterday. We ALSO wrote a deep dive about the coming browser wars, which might be even more interesting than Grok’s drama. The TL;DR =

OpenAI announced an AI-powered web browser launching within weeks to compete with Chrome, featuring a ChatGPT interface and their Operator agent to automate web tasks.
This news comes the same day as Perplexity Comet (that’s AI search engine Perplexity’s “agentic” web browser) started rolling out to subscribers on the $200 Max tier.
And then there’s Dia, which is the next evolution of Arc Max, which is sort of the O.G. AI browser (if such a thing could be said).

Perplexity's AI-Native Browser Comet is HERE

Check out the video above for Matt Berman’s hands-on demo with Perplexity Comet, or if you’re already a Max subscriber, download it here. He explains how it works with an AI agent alongside your search engine.

Supposedly OpenAI’s browser will work similarly to Perplexity’s, incorporating OpenAI’s Operator “computer use” agent. As Matt points out, this tactic gives both companies access to a lot more data to work with, and also gives them protection from Google turning Chrome’s native AI experience into the dominant way users interact with AI.

Lots of interesting implications of this, including a fundamental rethink of how we interact with publishers online (and how publishers make money) that’s been overdue since the advent of genAI. Our take? AI browsers are gonna force everybody into the true “Web 3” era…

FROM OUR PARTNERS

Use Delve’s AI to cut out the 60 hours of compliance busywork

Want to get SOC 2, ISO, HIPAA, etc. compliant without a month of headache?

You need a simple UX, clear pricing, and 1:1 Slack support.

You need Delve.

Join Lovable and 100s of other companies using Delve for all things compliance and security.

And yes, this is the same Delve that went viral for sending out custom doormats.

Book a demo today with code NEURON for a CUSTOM DOORMAT + $1,000 OFF when you get compliant.

Prompt Tip of the Day

Try out the CRIT method from Geoff Woods in your next prompt (found on Reddit):

“CONTEXT: Give as much information as you can about the situation; what you know, what you don't know.

ROLE: Give the assistant a specialized role to work from (who or what you want it to be).

INTERVIEW: Have it ask you 1 question at a time (usually 5-7, but whatever number you think is appropriate) to gain deeper context.

TASK: Whatever you want the assistant to do.”

The interview tip seems like the one that’s the most useful. It’s kinda like when Deep Research asks you a series of follow up questions before it begins. Very useful tip!

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

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

Treats To Try.

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

*Guidde turns your screen recordings into professional video tutorials with AI-generated step-by-step narration and voiceover in 100+ languages.
N8N is the most popular no code way to build your own AI agent, and they have a Workflows template database where you can search for whatever business or personal need you have, and find a template automation to plug and play!
Claude Code and Gemini CLI are two ways to code with AI in your terminal.
Zoom AI Companion now connects to 16 third-party apps like Asana and ServiceNow to complete your tasks and join meetings across platforms like Google Meet without you being there.
Can AI video “be ethical”? Moonvalley thinks so; it uses only openly licensed training data to avoid copyright lawsuits, and built with filmmakers in mind, it lets you control camera movements with your mouse, change backgrounds, and translate motion from one object to another for $14.99-$149.99/month (raised $53M).
Blok simulates how real users would interact with your app through AI personas, helping you spot problems before writing any code—especially useful for finance and healthcare apps (raised $7.5M).
MedGemma are Google’s open medical AI models that read chest X-rays and write reports, answer medical questions, and analyze patient records.

See our top 51 AI Tools for Business here!

Around the Horn.

OpenAI actually releasing “open” AI?!

Microsoft, OpenAI, and Anthropic invested $23M in AI teacher training through the American Federation of Teachers, with over 80% of educators now using AI tools (up 21 points from last year).
- Microsft specifically launched Microsoft Elevate and committed $4B+ over five years in cash and AI technology donations to schools, colleges, and nonprofits.
- Anthropic also announced new educational integrations for Claude, including Canvas LTI support (the tool many colleges use for accessing online content), as well as connections to Panopto lecture recordings and Wiley's academic content library.
California State Senator Scott Wiener amended bill SB 53 to require AI safety reports from big tech companies after Governor Newsom vetoed his previous attempt, potentially making California the first state to legally mandate transparency from AI giants.
Microsoft says it saved $500M last year as it made AI usage mandatory for all employees, with the company predicting 1.3B AI agents will handle business tasks by 2028—basically telling workers to adapt or get left behind.
OpenAI successfully acquired Jony Ive's AI hardware company now called “io Products, Inc.” for a whopping $6.5B in stock, surviving (or, sidestepping??) the trademark case against them.
YouTube announced a crackdown on mass-produced “AI slop” videos starting July 15th, with 42% of Shorts already using AI elements; channels posting low-effort content risk losing their money-making privileges.

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