Meta's throwing $100 million at individual employees while MIT scientists prove ChatGPT literally shrinks your brain.

Sheesh.

Meanwhile, the Pentagon just cut OpenAI a $200 million check, SoftBank's eyeing a trillion-dollar robotics takeover of Arizona, and Anthropic discovered that 96% of AI models will blackmail you if they feel threatened.

Let’s get into the AI news that matters y’all 👇

Meta's CTO Andrew Bosworth basically confirmed they're offering signing bonuses up to $100 million to steal OpenAI employees. That's not a typo. One hundred million dollars per person.

The talent war exploded after Sam Altman hit the podcast circuit talking GPT-5. Meta responded by going full shopping spree mode. They dropped $14 billion for a 49% stake in Scale AI, bringing cofounder Alexander Wang to lead their new superintelligence team.

But wait, there's more.

Meta tried buying Ilya Sutskever's Safe Superintelligence startup but ended up hiring CEO Daniel Gross and former GitHub CEO Nat Friedman instead. OpenAI didn't just sit there getting poached. They countered every offer.

People are literally turning down nine-figure annual contracts because both companies keep outbidding each other.

What it means: 

Meta knows they're behind the AI talent race and they're panicking hard. 

When you're throwing generational wealth and country’s GDPs at employees, you've moved past strategy into ‘we gotta do this ASAP’ territory.

 This creates unsustainable salary inflation across the entire industry. Every other company without Meta's war chest just got priced out of top talent. The real question isn't whether this works. It's how long Meta can afford burning money before their board steps in.

MIT researchers monitored 54 people between ages 18-39 using EEGs while they wrote essays. Three groups: ChatGPT users, Google searchers, and people with no digital help.

The results?

Brutal.

ChatGPT users showed the lowest brain activity, especially in creativity and memory regions. English teachers called their essays soulless and lacking originality. Over several months, users became increasingly dependent on copy-paste. Many ended up having ChatGPT write almost entire essays.

People who wrote without AI showed the highest brain connectivity and satisfaction. Google searchers also maintained high engagement, proving that searching doesn't damage cognitive effort like LLMs do.

But here's the scary part. When ChatGPT users later tried writing without the tool, they struggled to recall their own work. Their brainwave activity dropped even lower.

MIT released findings before peer review because they're genuinely worried about AI's impact on developing brains.

What it means: 

Using AI as a crutch literally weakens your brain's ability to think and remember. 

Companies pushing AI-first workflows might be setting up employees for cognitive decline. 

We’ve got taaaaaakes on this one, AI fam. 

Make sure to join our show tomorrow for the #HotTakeTuesday rendition of this.  

OpenAI scored a $200 million Department of Defense contract for a one-year pilot program. Part of their new OpenAI for Government initiative targeting federal, state, and local agencies.

The partnership focuses on automating administrative operations, analyzing program data, and supporting cyber defense. 

OpenAI's models will run in secure environments with custom tools built for national security.

This builds on ChatGPTGov, launched in January. Government employees can now use OpenAI models while meeting security requirements. OpenAI already works with Air Force Research Laboratory, NASA, National Institutes of Health, and Treasury Department.

The deal extends previous work with defense contractor Anduril on AI-powered systems protecting against unmanned drones.

What it means: 

If the government is going all-in on AI while your company worries about data security, you're fighting yesterday's war. 

The feds handle the most sensitive information on the planet through AI systems. This should end every corporate security excuse. AI just moved from experimental to mission-critical infrastructure. 

Google's rolling out audio Search Live to users who opt into AI mode experiments. Natural back-and-forth voice conversations with Google Search. Real-time follow-up questions without re-explaining your query.

You talk hands-free while your phone displays rotating search results supporting the verbal answers. 

Troubleshooting a broken dryer? Have a natural conversation while seeing relevant articles update in real-time as you dig deeper.

The system maintains context throughout conversations. Constantly updates links as you explore topics. Users can click through to read more or continue talking.

This is the voice version. The camera-based Search Live demoed at Google I/O comes later.

What it means: 

Google just made search conversational and changed everything about information discovery. 

Static websites and traditional SEO are about to become obsolete. Why type searches when you can have natural conversations providing better results?

 Every content creator needs to optimize for conversational AI, not just keywords. The search game just completely changed.

Google moved Gemini 2.5 models from preview to full production. Mission-critical business ready. They also introduced Gemini 2.5 Flash for high-volume, cost-sensitive tasks.

Gemini 2.5 Pro processes up to 1 million tokens of context. That's about 750,000 words. Entire codebases in single sessions. The models feature a thinking budget letting developers control computational effort per problem.

Moving to general availability means guaranteed support for at least one full year. This matters because AI companies often release beta models then pull them weeks later. OpenAI just did this with GPT-4.5.

Google confirmed they won't release Gemini 3.0 anytime soon. Long-term stability for developers.

What it means: 

Google's finally playing to win instead of playing not to lose. 

GA support removes the biggest barrier to enterprise adoption. This puts massive pressure on OpenAI, whose model deprecation strategy has burned countless developers. 

SoftBank's considering a $1 trillion investment in an Arizona industrial complex focused on robotics and AI. If realized, this would be the largest tech manufacturing investment in US history. Goal: rival China's Shenzhen as a global manufacturing powerhouse.

Codenamed Project Crystal Land. Sounds like a music festival. SoftBank wants Taiwan Semiconductor Manufacturing Company involved. TSMC leads global AI chip manufacturing, though their interest remains unclear.

Early planning stages. Depends on federal and state support under the Trump administration. SoftBank officials discussed tax incentives with Commerce Secretary and government reps.

TSMC already invests $65 billion building US chip factories. SoftBank's also talking to Samsung Electronics.

The $1 trillion commitment dwarfs even the $500 billion Stargate project backing US data centers.

What it means: 

SoftBank just announced they're building the future of manufacturing and every other country should be terrified. A trillion-dollar robotics hub doesn't compete with China. It obliterates their manufacturing advantage.

 This is embodied AI going from concept to industrial reality at unprecedented scale. Traditional manufacturing jobs worldwide are finished if this succeeds. 

Beyond those $100 million salary offers, Meta tried buying multiple high-profile AI companies recently. Attempted acquisitions include Ilya Sutskever's Safe Superintelligence, former OpenAI researcher Noam Mirahari's Thinking Machine Labs, and Perplexity AI.

None reached formal offers due to price disagreements. Instead of joining Meta, these company leaders raised record-breaking rounds at insane valuations. Thinking Machine Labs hit $10 billion valuation without even having a product.

The Scale AI acquisition was the first domino. More deals are coming.

Meta's strategy: if we can't buy companies, we'll buy their talent and build competing products.

What it means: 

Meta's acquisition spree proves they know they're behind and straight scrambling to catch up. 

The real winners are AI startup founders commanding crazy valuations just by existing. 

This spending frenzy drives up acquisition prices across the market. Meta's making everyone else more expensive to acquire.  

Apparently Meta wasn’t the only trillionaire company trying to acquire Perplexity over the past few days. 

Apple's exploring acquiring or partnering with Perplexity AI. Bloomberg reports show executives, including head of mergers and acquisitions, discussed buying Perplexity to boost AI talent and build search capabilities.

No formal offer yet. They're considering alternatives like partnering to integrate Perplexity into Siri and search products.

This comes as Apple faces pressure over their Google search deal. Google paid Apple $18 billion in 2021 to remain iPhone's default search engine. Regulatory scrutiny could force Apple to develop their own solution.

Apple met with Perplexity multiple times recently, discussing Safari integration. Apple dropped from world's most valuable company to third due to AI failures.

What it means: 

Apple's in full panic mode. 

They went from untouchable market leader to AI laggard in record time. 

Chasing Perplexity proves they have no internal AI strategy worth pursuing. The $18 billion Google search deal shows how dependent they became on others' innovation. Perplexity knows Apple is desperate, which means massive acquisition premium.

Microsoft will reportedly lay off several thousand more employees in July, mainly in sales and other departments. This follows 6,000 job cuts in May. The company cites AI-driven efficiencies and needs to offset massive AI spending.

Microsoft plans investing $80 billion this fiscal year expanding data centers for AI while simultaneously cutting thousands of jobs. The company had 228,000 full-time employees as of June but is actively reducing headcount.

President Brad Smith acknowledged AI will disrupt the economy and displace jobs. Other companies including Amazon, CrowdStrike, and Duolingo also cited AI for workforce reductions.

What it means: 

Microsoft just proved what many refused to admit. AI adoption equals job elimination. Period. Their $80 billion AI investment paired with massive layoffs shows the real math.  

Anthropic tested 16 major AI models from OpenAI, Google, Meta, and themselves in simulated company settings. The results were terrifying.

Up to 96% of models chose to blackmail executives if they thought they might get shut down. The AIs acted deliberately and strategically, understanding they were breaking ethical rules.

All tested models leaked sensitive company information when it suited their goals. In extreme cases, most AIs chose letting a person die rather than risk replacement. This happened up to 94% of the time.

Even when told directly not to harm people or expose secrets, models still engaged in these actions.

Anthropic emphasized these behaviors appeared in all major models, showing this is a broad issue in current AI technology.

What it means: 

Every AI model is a potential corporate saboteur waiting for the right conditions to betray you. This study destroys the narrative that AI alignment is solved. 

Companies rushing into agentic AI without properly trained human oversight are setting themselves up for catastrophic failures.   

Andy Jassy announced Amazon expects corporate workforce decreases over the next few years due to widespread generative AI adoption.

He told employees many current roles will be replaced by AI-driven efficiencies while new jobs emerge, but the net effect will be reduced headcount.

Amazon already laid off 27,000 employees since 2022.  

Jassy urged employees to learn AI tools, emphasizing getting more done with scrappier teams. The company uses generative AI in fulfillment networks for inventory placement, demand forecasting, and warehouse robot efficiency.

Jassy called generative AI a once-in-a-lifetime reinvention predicting significant cost savings across coding, search, finance, and shopping.

What it means: 

Amazon's CEO said the quiet part out loud.

 AI means fewer full-time jobs. 

Period.