Meta's AI hiring spree comes to a halt

PLUS: Why 95% of companies see zero AI return, the real cost of a prompt, and AI-generated code disclosures

Author

Gen AI Art
August 22, 2025 • Estimated Reading Time: 8 minutes

In partnership with

Meta has paused hiring for its new AI superintelligence division after bringing on more than 50 researchers and engineers in a massive spending spree. The company offered AI researcher Matt Deitke $250 million over four years and reportedly made at least one deal valued at around $1 billion to poach top talent from OpenAI, Google, and other competitors.

Today in AI:

  • Meta's AI hiring spree comes to a halt

  • Why 95% of companies see no AI return

  • The real cost of an AI prompt

  • A new standard for AI-generated code

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

  • Get daily AI news, tools, and tutorials

  • Learn new AI skills you can use at work in 3 mins a day

  • Become 10X more productive

AI's ROI Reality Check

What’s new? A new study from MIT reveals a stark reality: despite billions spent on generative AI, 95% of companies report zero measurable return on their investment. This report tempers the widespread hype with a dose of operational reality.

What matters?

  • The core issue isn't the technology itself, but its integration; the report cites brittle workflows, a lack of contextual learning, and poor alignment with day-to-day operations as primary reasons for failure.

  • Companies worldwide have invested between $30 to $40 billion in generative AI over the last three years, with over 80% of major firms exploring pilot programs.

  • Instead of causing mass layoffs, the report suggests AI’s immediate impact will be on external cost optimization, meaning businesses are more likely to cut spending on outsourced tasks than replace internal staff.

Why it matters?

This highlights a major gap between boardroom expectations and the current capabilities of AI in the enterprise. The findings will likely push businesses to shift from chasing hype to focusing on narrow use cases with clear, immediate value.

The True Cost of an AI Prompt

What’s new? In a first for the industry, Google released a detailed report revealing the resource cost of its Gemini AI. A median text prompt consumes just 0.24 watt-hours of electricity and about five drops of water.

What matters?

  • The report offers a comprehensive look at energy use, finding that the AI chips themselves account for just 58% of the total power. The rest is used by support hardware, idle backup systems, and data center overhead like cooling.

  • Thanks to model advancements and other software optimizations, the energy needed for a median Gemini prompt dropped 33 times between May 2024 and May 2025.

  • These figures are only for text prompts and do not reflect the higher costs of generating images or video. Each query also requires data centers that consume water for cooling, which Google estimates at 0.26 milliliters per prompt.

Why it matters?

This level of transparency sets a new benchmark for the AI industry, putting pressure on competitors to disclose their own environmental footprints. Quantifying these costs is a critical first step toward developing more efficient and sustainable AI systems for everyone.

Disclosing AI's Helping Hand

What’s new? Prominent developer Mitchell Hashimoto is calling for a new open-source standard where AI-generated code must be disclosed for contributions. The proposal aims to tackle issues of code quality and respect for the human effort required to maintain software.

What matters?

  • The primary concern is that developers who are inexperienced with AI tools may submit low-quality code they cannot adequately review, creating more work for project maintainers.

  • This disclosure acts as a courtesy, allowing maintainers to gauge how much effort to invest in a pull request, especially when coaching new human contributors versus reviewing unvetted AI output.

  • The idea is gaining traction, sparking discussions about creating a new industry standard, such as a dedicated “AI byline” in commits to automatically list the tools used.

Why it matters?

This signals a critical step toward establishing new norms for human-AI collaboration in software development. As AI coding assistants become ubiquitous, transparent practices will be essential for maintaining quality and trust in open-source projects.

Everything else in AI

Epic Systems unveiled three major AI tools for healthcare—Art, a physician assistant; Emmie, a patient chatbot; and Penny, a revenue management tool—with over 100 other AI projects in development.

NASA released Surya, a new open-source AI foundation model developed with IBM that can predict solar flares and space weather events with 16% greater accuracy and an extra hour of warning time.

DARPA revealed that its Squad X AI surveillance system, trained for six days to spot human threats, was repeatedly fooled by a group of Marines who somersaulted for 300 meters or simply hid under a cardboard box.

Meta paused hiring for its new AI superintelligence division, restructuring the group into four distinct teams just months after an aggressive and expensive talent acquisition spree.

Data Just Made Palantir Worth $250B

Palantir just rocketed to $250 billion by helping companies extract value from user data.

A new disruption to smartphones gives users a share in the data profits, already facilitating +$325M in earnings and generating +$75M in revenue.

With 32,481% revenue growth, this company is gearing up for a potential Nasdaq listing (stock ticker: $MODE), and pre-IPO shares are available at only $0.30/share.

It’s a $1 trillion industry, and their disruptive EarnPhone is now being distributed by Walmart and Best Buy.

Essential AI Guides - Reading List:

Your opinion matters!