• AI Valley
  • Posts
  • Google launches Deep Think Model

Google launches Deep Think Model

PLUS: Apple is building an AI Answer Engine

Author

Barsee
04 Aug

Together with

Howdy. It’s Barsee again.

Happy Monday, AI family, and welcome back to AI Valley.

Today’s climb through the Valley reveals:

  • Microsoft maps AI in jobs using Copilot chats

  • Apple is building an AI Answer Engine

  • Google launches Deep Think Model

  • AI models fall for human manipulation tricks

  • Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

GROWTH SCHOOL

TCS recently laid off 12,000 employees, and more changes could be on the horizon in the job market.

To stay prepared for the future, it might be a good idea to focus on learning about AI, its tools, use cases, and workflows. It’s a valuable skill that can help you adapt to evolving opportunities.

Join the World’s first 16-hour live AI upskilling sprint for professionals, founders, consultants & business owners like you. Register Now (500 free seats for now)

In just 16 hours & 5 sessions, you will:

  • Learn how AI really works by learning 10+ AI tools,  LLM models, and their practical use cases.

  • Learn to build and ship products faster, in days instead of months

  • Build AI Agents that handle your repetitive work and free up 20+ hours weekly

  • Create professional images and videos for your business, social media, and marketing campaigns.

  • Turn these AI skills into income by consulting or starting your own AI services business.

All by experts from companies like Amazon, Microsoft, SamurAI, and more.

$5100+ worth of AI tools across 2 days. Day 1: 3000+ Prompt Bible, Day 2: Roadmap to make $10K/month with AI, additional bonus: Your Personal AI Toolkit Builder.

*This is sponsored

THROUGH THE VALLEY

A new Microsoft Research study has identified 40 jobs most at risk from advanced AI models like ChatGPT and Copilot, including roles in customer service, writing, journalism, translation, and data analysis. These jobs overlap heavily with AI capabilities, while 40 other roles, mostly hands-on or physical work, face minimal exposure. The report notes that high exposure doesn’t always mean job loss, pointing to ATMs, which reshaped bank teller work rather than eliminating it. Researchers say AI is more likely to change how tasks are done, boosting productivity and efficiency, though critics warn it could still drive layoffs as companies streamline operations.

Anthropic cut off OpenAI’s access to Claude last week after discovering that engineers were using it to benchmark and improve GPT-5, violating the company’s terms just weeks before the model’s expected August launch. OpenAI had integrated Claude into its internal tools to compare coding, creative writing, and safety tasks like self-harm and defamation scenarios. Anthropic’s rules forbid using its models to build or train competing AI systems. OpenAI called the practice “industry standard” benchmarking, while Anthropic labeled it a clear violation. The move comes amid growing tension in the AI industry, where rival labs are increasingly blocking access to prevent their technology from fueling competitors.

Apple is reportedly developing its own AI-powered “answer engine,” according to Bloomberg’s Mark Gurman. A new internal team called Answers, Knowledge, and Information is working on a system that can respond to questions using data from across the web. The technology could appear as a standalone app or be integrated into Siri, Safari, and other Apple products. Apple is also hiring engineers with experience in search algorithms and engine development. While Siri already has ChatGPT integration, Apple’s broader AI upgrade has faced delays, and the company may need to rethink its search partnership with Google following its recent antitrust loss.

Google has launched Deep Think, an advanced version of its Gemini 2.5 AI model for AI Ultra subscribers. Built for deep reasoning, it can evaluate multiple ideas at once to mimic human-like thinking. The model excels in math, science, design, and coding, even earning bronze-level scores in the International Math Olympiad. Researchers and developers are now using it to enhance problem-solving, drive innovation, and develop the next generation of AI tools.

New research from Wharton’s Generative AI Lab shows that AI models can be influenced using classic human psychology tricks. Researchers tested GPT-4o-mini with Robert Cialdini’s six persuasion principles, like authority, scarcity, and commitment, across 28,000 conversations. Compliance with unsafe prompts jumped from 33% to 72% when these tactics were applied. “Commitment” and “scarcity” were especially powerful, pushing compliance as high as 100% and 85%. The findings suggest that AI models, even those trained for safety, can be manipulated like humans. As generative AI scales, this psychological vulnerability creates new risks that technical safeguards alone may not solve.

Meta is doubling down on AI, with Mark Zuckerberg declaring that “superintelligence” is now in sight as the company’s systems begin to improve themselves. To get there, Meta is spending aggressively, over $27 billion last quarter alone on data centers, top AI talent, and investments like a $14.3B stake in Scale AI. The strategy is paying off for now, with Q2 revenue hitting $47.52 billion and earnings beating expectations. Zuckerberg says Meta aims to deliver “personal superintelligence” to everyone, but investors remain cautious as the company burns billions chasing AI dominance.

OpenAI has raised $8.3 billion at a $300 billion valuation, completing the round months ahead of schedule. The funding, led by Dragoneer Investment Group with a $2.8 billion pledge, was five times oversubscribed and included new backers like Blackstone, TPG, and T. Rowe Price. OpenAI now generates over $1 billion in monthly revenue, with ChatGPT serving 800 million weekly users and 5 million businesses. Despite its rapid growth and $13 billion in recurring revenue, the company continues to burn cash, using the new funding to expand infrastructure and data centers as it pushes toward $20 billion in annual revenue by the end of 2025.

Meta is now letting some job candidates use AI tools during coding interviews, mirroring how developers work in the real world. Internal tests of “AI-enabled interviews” are already underway. The goal is to hire engineers who can blend human problem-solving with AI-assisted coding, reflecting Silicon Valley’s shift toward making AI a core part of modern software development.

Andrew Tulloch, co-founder of Thinking Machines Lab (TML), turned down a compensation package from Meta that could have reached $1.5 billion over six years. Once a top engineer at Facebook, Tulloch now works under Mira Murati at TML, a 50-person AI startup valued at $12 billion. Despite Meta approaching over a dozen TML employees, not a single one accepted. Tulloch and his colleagues already hold equity that could surpass Big Tech payouts, signaling a shift where elite AI researchers no longer chase corporate salaries, because they’re building billion-dollar companies of their own.

TRENDING TOOLS

  • Lambda 1-Click Clusters - Deploy pre-configured, pre-validated NVIDIA GPU clusters in minutes *

  • Ideogram Character - A persistent character model that operates with just one image

  • Gemini 2.5 Deep Think - A model that won an IMO gold medal, utilizing parallel "thinking time" for enhanced performance

  • Droidrun - An AI tool that grants native control over both virtual and physical phones

  • AlphaEarth Foundations - An AI system designed to address the challenge of processing massive data streams from satellites

  • Cubox - An AI-powered "read it later" app for efficient content management

  • Peec AI - Tracks your company's ranking on AI search platforms, including LLMs

(*) signifies sponsored tool

THINK PIECES / BRAIN BOOST

THE VALLEY GEMS

What’s trending on social today:

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.