☕️ Google employee charged for insider trading on Polymarket

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🚨 Google employee charged for insider trading on Polymarket

🍎 Leak reveals Apple’s plans to take on ChatGPT and more

☁️ Meta might launch a cloud business

📱 Meta launches Instagram, Facebook, and WhatsApp subscriptions

🤖 Tesla starts Optimus factory build at Giga Texas

Plus: 🎁 12 other news you might like, 🧰 6 tools, and 📚 4 papers.

🚨 Google employee charged for insider trading on Polymarket LINK

Federal prosecutors have charged Michele Spagnulo, a 36-year-old Google software engineer based in Switzerland, with making more than $1.2 million on Polymarket by betting with confidential internal Google search data.

Trading under the username "AlphaRacoon" between October and December, Spagnulo correctly wagered that singer D4vd would be Google's most-searched person of 2025, an outcome the market had given a near-zero probability.

Spagnulo faces counts of commodities fraud, wire fraud, and money laundering, and Google says he accessed marketing material through a tool open to all employees before being placed on leave.

🍎 Leak reveals Apple’s plans to take on ChatGPT and more LINK

Apple is preparing a standalone Siri app for iOS 27, set to compete directly with chatbots like ChatGPT, Claude, and Gemini, according to leaked renders published by Bloomberg ahead of June's Worldwide Developers Conference.

The Siri app will show your past chat history and let you upload documents and photos alongside text, while a rebuilt AI model behind Siri uses Google's Gemini technology under the hood for added intelligence.

Siri will also weave into iOS 27 through the Dynamic Island for button-triggered queries, and swiping down for Spotlight Search will now pull AI-powered results shown in a card-style interface for launching apps, messages, and shortcuts.

☁️ Meta might launch a cloud business LINK

Meta is considering getting into the cloud computing business, with CEO Mark Zuckerberg telling Wall Street the company could lease out some of its computing resources if it finds itself with more capacity than needed.

Zuckerberg said many companies reach out to Meta every week asking for API services or offering to pay a premium for compute, though Meta hasn't sold any yet because it still has its own uses for the resources.

The remarks follow Meta's April decision to raise its 2026 AI capital expenditure forecast to between $125 billion and $145 billion, up from the earlier range of $115 billion to $135 billion.

📱 Meta launches Instagram, Facebook, and WhatsApp subscriptions LINK

Meta is rolling out paid "Plus" tiers for Instagram, Facebook, and WhatsApp today, after months of quiet testing, giving subscribers access to features that free users on the three apps simply can't reach.

Instagram Plus and Facebook Plus each run $3.99 a month while WhatsApp Plus costs $2.99, with perks like detailed Story stats, extended vanishing posts past 24 hours, custom themes, and exclusive reactions.

Instagram Plus subscribers can pick multiple audiences for Stories, see who rewatched, search viewer lists, make "spotlight" Stories, send a "super heart," watch Stories invisibly, and post without showing up in followers' feeds.

🤖 Tesla starts Optimus factory build at Giga Texas LINK

Tesla has started building its dedicated Optimus factory at Gigafactory Texas, with the first steel structure now standing on the North Campus and the second phase of land reclamation moving forward, according to drone footage from May 27.

The new building will stretch nearly the full length of the main Giga Texas factory, potentially over 4,000 feet, and sits within a North Campus expansion adding more than 5.2 million square feet of industrial space.

Tesla plans to build about 10 million Optimus robots a year at the site, roughly 27,000 units each day, with a second-gen production line targeting high-volume output starting in Summer 2027.

Other news you might like

• Snowflake soars 33% on earnings beat and plan to spend $6 billion on Amazon cloudLINK

• Qualcomm promises $300 Windows laptops with new Snapdragon CLINK

• EU fines Temu €200m for failing to prevent the sale of illegal productsLINK

• FCC filing shows Amazon plans to acquire Apple’s 20% stake in GlobalstarLINK

• Mistral to explore designing own chips, CEO says, as it ramps up infrastructure buildLINK

• I think Anthropic and OpenAI have found product-market fitLINK

• New report shows annual app subscribers rarely return after they cancelLINK

• Payroll startup Remote says it grew revenue 50% per employee without adding headcountLINK

🧰 Trending tools

SpotsNow: tracks podcast ad spend, reveals competitor campaigns, and lets you purchase available inventory directly from publishers in one platform.LINK

Pitch Agent: an AI-powered presentation tool that helps teams collaboratively build on-brand slide decks with analytics to track client engagement.LINK

Revolte: an AI-powered platform that automates code generation, security checks, PR creation, and deployment monitoring while keeping engineers in control of key decisions.LINK

Angel Match 4.0: a searchable database of 125k+ investors with filters for focus, location, and past investments, saving founders hours of manual research.LINK

Robinhood Agentic Trading: lets you trade tokenized versions of 2,000+ stocks and ETFs as derivatives, enabling price exposure without direct asset ownership.LINK

Buffer API: lets you schedule and manage social media posts across 10 platforms programmatically, with MCP server, CLI, and no-code automation support included.LINK

📚 Trending papers & reports

FuzzingBrain V2 is a multi-agent AI system that automatically finds and confirms real software vulnerabilities, achieving a 90% detection rate across 40 bugs and discovering 29 zero-day vulnerabilities in real-world deployment.LINK

Gemini Embedding 2 is a single AI model that converts video, audio, images, and text into a shared representation space, beating specialized models on key benchmarks including a score of 84.0 on MTEB Code.LINK

GitHub Actions workflows from 260K real repositories show that larger, more complex configurations fail more often and require more maintenance, with only a small set of features used frequently.LINK

AI agents can now teach themselves new skills and store them in memory, allowing the system to grow smarter over time without human retraining.LINK

Want to get the latest news differently? Find us on:

See you tomorrow for a new dose of ☕️ Techpresso!