Build AI That Acts: Google Cloud Hackathon

Move from insights to execution. Join the Google Cloud Rapid Agent Hackathon to build autonomous AI agents that solve real-world enterprise challenges.

Harness Google Cloud to build agents that don't just chat—they act. Integrate with key ISVs to automate complex workflows and drive real business value. Showcase your technical excellence, solve pressing enterprise problems, and lead the charge in the agentic AI revolution. Join a global community of developers building the next generation of intelligent tools.

Register Today. Lead the Agentic AI Revolution!

Register Now →

🔍 Billions Pour Into Post-Google Search

Google's announcement to replace its traditional search interface triggered a wave of investment in search startups. Exa Labs, backed by Andreessen Horowitz, raised $250 million at a $2.5 billion valuation. Parallel Web Systems, led by former Twitter CEO Parag Agrawal, raised $100 million at a $2 billion valuation from Sequoia Capital. Tavily and TinyFish are also competing for the same market.

Platforms like Amazon, LinkedIn, and Reddit are revamping their own search features with AI.

Key Insights:

• Exa Labs leads the raise - Andreessen-backed Exa Labs raised $250 million at a $2.5 billion valuation to compete directly in AI-powered search

• Parallel Web Systems enters at scale - Former Twitter CEO Parag Agrawal's startup raised $100 million at a $2 billion valuation in a Sequoia-led round

• A broader wave forming - Exa, Tavily, TinyFish, and Parallel are each building distinct approaches to next-generation search

• Platforms joining the race - Amazon, LinkedIn, and Reddit are rebuilding their search features with AI, creating potential acquirers for startups

• ChatGPT held the prior layer - Before Google's announcement, ChatGPT handled the majority of AI-powered searches daily, but OpenAI cannot prioritize search while also racing on model development

The Bigger Picture: Google's move to AI search confirmed the opportunity is real. For years, investors treated AI search as too close to Google's core turf. Now that Google has declared the old model obsolete, focused companies may carve out durable niches. The smart money at Sequoia and Andreessen thinks so. Startups that find product-market fit become attractive acquisitions for platforms like Amazon and Reddit that want AI search capability but do not want to build it.

Read More →

🚗 Uber Returns to Self-Driving Roads

Uber is deploying sensor-equipped vehicles on its rideshare network through a new program called AV Lab. The cars, starting with a single Hyundai Ioniq 5, will be manually driven but carry cameras, lidar, and radar to collect driving data.

That data will be shared for free with robotaxi partners including Wayve, WeRide, Nuro, and Waabi. The project addresses a data shortage among partners who lack the resources of larger developers like Waymo.

Key Insights:

• Scaling fast - Uber expects the AV Lab fleet to generate at least 2 million miles of driving data per month by year-end, scaling further in 2027

• Free data for partners - All driving data will be provided at no cost to AV partners, addressing a critical resource gap for smaller robotaxi developers

• Network advantage - Uber's rideshare network handles 40 million trips daily, exposing the fleet to edge cases autonomous vehicles must master

• Milestone threshold - CFO Balaji Krishnamurthy estimates AV operators need at least 10 million miles of data before a public driverless launch

• A cautious return - Uber sold its AV division in 2020 following a fatal accident in Tempe, Arizona, and has since focused on partnerships over direct development

The Bigger Picture: Most robotaxi startups are burning cash without the scale to collect the data they need. Waymo can run thousands of vehicles across multiple cities. Most of Uber's partners cannot. By deploying a data collection fleet on its existing network, Uber becomes the central infrastructure layer for the autonomous vehicle industry - solving its partners' biggest constraint rather than competing with them. That is what platform scale looks like in practice.

Read More →

🔢 AI Cracks an 80-Year Math Mystery

An OpenAI reasoning model has disproved a long-standing conjecture in discrete mathematics, resolving a problem Paul Erdos first posed in 1946. The question, how many pairs of points in a plane can be exactly distance one apart, had stood open for nearly 80 years. The model was a general-purpose reasoning system, not one trained for mathematics. Its proof was reviewed and confirmed by external mathematicians, who also wrote a companion paper.

Key Insights:

• 80 years open - The planar unit distance problem was posed by Paul Erdos in 1946 and considered one of the best-known unsolved problems in combinatorial geometry

• General-purpose model - The proof came from a broad reasoning model as part of a test of whether frontier AI can do original research

• Externally validated - A team of outside mathematicians independently checked the proof and wrote a companion paper on its significance

• Unexpected technique - The model applied ideas from algebraic number theory to an elementary geometric question - a pairing reviewers called sophisticated and unexpected

• Top-tier endorsements - Fields medalist Tim Gowers called it a milestone in AI mathematics, and number theorist Arul Shankar said AI now has original ingenious ideas

The Bigger Picture: This goes far beyond a single math problem. For decades, AI's role in science has been to assist, processing data, identifying patterns, suggesting next steps. This proof shows autonomous discovery: AI generating an original result on an unsolved problem without being guided to the answer. If this extends to other domains, it reshapes timelines on drug discovery and materials science. When Fields medalists call a result a genuine milestone, it is one.

Read More →

🎬 Hollywood Finds AI's Real Role

AI video companies have pivoted from pitching studios on camera replacement to integrating into full production workflows. Luma AI CEO Amit Jain says typical AI clips run 10 to 16 seconds, too short to be a shot or scene. The company now targets studio workflow integration over standalone clip generation.

Read More →

🔬 Google Builds AI for Research Labs

Google launched Gemini for Science at I/O 2026, a suite of three AI research tools. The collection includes a hypothesis generator running multi-agent idea tournaments, a computational discovery engine generating thousands of code variations in parallel, and a literature analysis tool built on NotebookLM.

Read More →

🔎 Google Rewrites Search After 25 Years

Google retired its traditional search box at I/O 2026, launching a redesigned interface that accepts text, images, PDFs, videos, and Chrome tabs. Head of Search Liz Reid called it the biggest upgrade in 25 years. The redesign merges AI Overviews and AI Mode into a single search flow.

Read More →

IrisGo 🖥️

An AI desktop companion that watches your workflows and learns to automate them. IrisGo, backed by Andrew Ng's AI Fund with a $2.8 million seed round, records any task once and handles it independently going forward. Built-in workflows include email drafting, invoice processing, and document summarization. Most data is processed on-device for privacy.

Try it →

Gemini Omni 🎬

Google's Gemini Omni Flash model creates and edits video from any combination of inputs - text, images, audio, or existing footage. Every instruction builds on the last, keeping characters consistent and physics intact across multiple edit rounds. Now rolling out to the Gemini app, Google Flow, and YouTube Shorts.

Try it →