🗞️In this edition

• Google upgrades Gemini with 3 flash

• Sponsored: Remio - AI powered personal knowledge hub

• Tech giants commit $500B to leases

• AI giants battle for India's users

• In other AI news –

  • Adobe under fire for allegedly using pirated books to train AI 

  • Gemini now lets users build mini apps with Opal

  • 4 must-try AI tools

Google upgrades Gemini app and search with faster, smarter 3 flash model

What's happening:

Google just made Gemini 3 Flash the default model inside its Gemini app and Google Search AI Mode. It's replacing Gemini 2.5 Flash.

The new model delivers "Pro-grade reasoning with Flash-level latency" meaning you get smarter answers without waiting longer. It outperforms last-gen flagship Gemini 2.5 Pro at "a fraction of the cost."

Tulsee Doshi, head of product at Google DeepMind, calls it a "huge upgrade." Faster responses. More detailed, nuanced answers. The model can analyze video, images, and text simultaneously, then generate plans in seconds.

Gemini 3 Flash is also rolling out to developers via Google AI Studio, Gemini API, Vertex AI, Android Studio, and more.

Why this is important:

Google's compressing flagship performance into a fast, cheap model.

That's the race right now who can deliver frontier reasoning without burning compute budgets or making users wait. 

OpenAI's doing it with o3-mini. Anthropic's doing it with Haiku. Google's betting Flash closes the gap.

The real move? Making it default in Search. That's billions of queries getting routed to smarter AI without users opting in.

Developers win too. Pro-level reasoning at Flash pricing changes unit economics for anyone building AI products.

Our personal take on it at OpenTools:

Google's playing the long game and most people are missing it.

They're not chasing flashy demos. They're making their best reasoning model the default experience for billions of searches. No opt-in. No friction. Just better AI baked into the product people already use daily.

The cost efficiency matters more than the performance benchmarks. Anthropic and OpenAI are burning cash to train bigger models. Google's squeezing flagship intelligence into a model that runs cheap and fast.

Developers should pay attention to the pricing arbitrage. If Gemini 3 Flash truly matches 2.5 Pro quality at Flash pricing, that's a 10x cost improvement for anyone routing high-volume requests.

The quiet winners here? Google Cloud and anyone building AI products who needs to stretch their inference budgets. The models are commoditizing fast efficiency is the new moat.

Read More

remio 2.0 - Your Personal ChatGPT

Tired of scattered notes, lost files, and forgotten conversations? 

remio unifies your digital life, transforming your knowledge into a personal intelligence source.

Here’s what you’ll love about remio:

• Tailored Answers: remio provides unique answers by combining with your personal knowledge.

• No More Uploads to ChatGPTs: One click to sync all your files, making your entire knowledge base chatable with AI.

• Master Your Meeting: Unlimited free recording with transcription, get AI summaries with key decisions.

• Privacy & Security: With a "Local First" design, all your data is stored exclusively on your device.

Get your personal ChatGPT now!

Oracle, Microsoft, and Meta lock in $500 billion worth of data center leases

What's happening:

Oracle, Microsoft, and Meta just committed $500 billion to data center leases over the next decade-plus. That's half a trillion dollars locked into server farms before they're even built.

Oracle leads the pack with $248 billion in commitments $150 billion signed in just three months. They're building infrastructure to power OpenAI's latest models under a $30 billion annual deal.

Meta's leasing commitments tripled to $58 billion year-over-year. Google's spiked 7x to $42.6 billion. Microsoft started early in 2023 to handle OpenAI's compute demand.

These obligations don't hit balance sheets until payments start. Leases can run 19 years. Renting avoids upfront development costs Meta's flagship Louisiana AI data center is being built by a special purpose vehicle taking on tens of billions in debt.

Wall Street's nervous. After Oracle disclosed its plans, investors sold the stock. DA Davidson analyst Gil Luria: "Oracle will be significantly strapped for capital."

Why this is important:

The AI infrastructure bet is now contractually locked in for two decades.

These aren't reversible capex decisions. They're long-term lease obligations that companies can't easily walk away from. If AI demand doesn't materialize or shifts faster than expected, billions are already committed.

Oracle's exposure is extreme. Their cloud business is small compared to AWS, Azure, and Google Cloud but they've committed more than anyone. The $30B/year OpenAI deal gives them "massive single-customer exposure." If that relationship sours or OpenAI's trajectory changes, Oracle's stuck with empty data centers.

The rent-vs-build strategy is a huge shift. Historically, hyperscalers owned their infrastructure. Now they're offloading construction risk to landlords and SPVs while locking in long-term payments.

Our personal take on it at OpenTools:

Oracle's either brilliant or reckless there's no middle ground here.

$248 billion in lease commitments for a company with $13 billion in annual free cash flow is wild. They're betting everything on AI demand staying hot for 15+ years. That's longer than the iPhone has existed.

The single-customer risk is what keeps us up at night. OpenAI's their anchor tenant at $30B/year, but OpenAI's burning cash and hasn't proven they can be profitable long-term. If that deal unravels, Oracle's holding the bag on hundreds of billions in leases.

Meta and Google are hedging smarter. They're renting AND building, spreading risk across owned and leased infrastructure. Microsoft moved early in 2023 they saw this coming and locked in capacity before prices spiked.

The rent-vs-own debate matters for one reason: flexibility. Owned data centers are sunk costs but you control them. Leased data centers are someone else's problem to build but you're locked in for decades.

Two years from now, someone's eating a $100B+ write-down on stranded data center capacity. Our money's on Oracle being first.

Read More

AI giants roll out Freebies in India to harvest multilingual training data

What's happening:

OpenAI, Google, and Perplexity just started an AI land grab in India giving away hundreds of dollars in subscriptions for free.

Google's offering its $400 Gemini AI Pro free for 18 months to 500 million Reliance Jio customers. OpenAI made ChatGPT Go (normally $54/year) free for a year only in India. Perplexity gave its $200/year Pro plan free to Airtel users.

The results are massive. ChatGPT's daily active users in India spiked 607% year-over-year to 73 million more than double the U.S. Gemini hit 17 million daily users vs. 3 million in the U.S. India's now the biggest market for both chatbots.

Perplexity went from 7% to over a third of its global daily users coming from India.

Why this is important:

India's 730 million smartphones and dirt-cheap data ($0.09/GB) make it the perfect testbed for AI adoption at scale.

But the real play isn't market share. It's data.

Five AI analysts told Reuters the freebies are about harvesting multilingual training data. Indian users mix languages mid-sentence, blend dialects, and code-switch constantly. That's exactly what current AI models struggle with and exactly what companies need to train on.

India's linguistic complexity is a "critical stress test" for AI models, according to Future Shift Labs co-founder Sagar Vishnoi. The free plans "fill gaps in AI training datasets that currently lack information on user behavior patterns in the region."

This is the Reliance playbook. Mukesh Ambani's telecom unit gave away free data and voice in 2016, ballooned to 500 million users, then turned on paid plans once everyone was hooked.

Our personal take on it at OpenTools:

This isn't charity. It's a data harvest disguised as market expansion.

OpenAI and Google aren't giving away $400-$600 annual subscriptions out of generosity. They're buying access to the most linguistically diverse user base on earth at scale.

Indian users generate training data that doesn't exist anywhere else. Code-switching between Hindi, English, Tamil, and regional dialects mid-conversation? That's gold for multimodal models trying to handle real-world complexity.

The numbers prove it's working. ChatGPT's 46% daily engagement rate in India crushes Perplexity (20%) and Gemini (14%). When people use AI three hours a day for free, companies collect millions of hours of behavioral data on how humans actually interact with AI.

The Reliance comparison is perfect. Free until you're hooked, then flip the switch. Except this time, the product isn't telecom it's your conversational patterns, language preferences, and how you think.

India's about to train the next generation of AI models whether they realize it or not.

Read More

📢 What else is happening?

• Adobe hit with proposed class-action, accused of misusing authors’ work in AI training – A proposed class-action lawsuit filed on behalf of Elizabeth Lyon, an author from Oregon, claims that Adobe used pirated versions of numerous books  including her own  to train the company’s SlimLM program.

• Google’s vibe-coding tool Opal comes to Gemini – The company on Wednesday said it is integrating the tool, which lets you build AI-powered mini apps, inside the Gemini web app, allowing users to create their own custom apps, which Google calls Gems.

👩🏼‍🚒Discover mind-blowing AI tools

1. Wave.video - Live streaming studio, video editor, thumbnail maker, video hosting, video recording, and stock library combined in one platform

2. Verbalate - A video translation and lip sync software designed to help businesses reach a global audience

3. SlangThesaurus - Allows you to effortlessly turn basic text into trendy internet slang, with customizable slang levels from 1 to 5

4. ChefGPT - An AI-powered recipe recommendation tool that suggests recipes based on the ingredients and tools you have