Friday 20 February 2026 | Join Free | Upgrade

Together with

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

💥 Ex-Google engineers charged with stealing secrets for Iran

🤖 AWS AI coding tool caused two major outages in December

🛡️ Google blocked 1.75 million risky apps in 2025

📱 Meta's metaverse is going mobile-first

💰 Nvidia nears $30 billion investment in OpenAI

💬 Meta shuts down Messenger standalone website

🎁 + 19 other news & articles you might like

🧰 + 4 trending tools

📚 + 2 trending papers & reports

FROM OUR PARTNER


🚀 What if your time tracking actually worked inside Jira?

Say goodbye to disconnected systems and manual headaches

Join leaders from TransUnion on March 11th for a live webinar to discover how they:

  • ✨ Simplified approvals
  • 📉 Reduced operational overhead
  • 👀 Gained true executive visibility
  • 💰 Saved nearly $1M annually with Tempo

Hear their real transformation story - moving from legacy tools to streamlined delivery.

Don't miss this inside look at how smarter time tracking can power smarter decisions.

👉 Reserve your spot today
💥 Ex-Google engineers charged with stealing secrets for Iran LINK
  • Federal prosecutors charged three former Silicon Valley engineers with stealing chip security trade secrets from Google and other companies and sending confidential files to unauthorized locations, including Iran.
  • Samaneh Ghandali allegedly transferred hundreds of Google files to a third-party platform, photographed trade secret information the night before traveling to Iran, and later signed a false affidavit denying it.
  • Google's internal security systems detected suspicious activity in August 2023, and all three defendants now face conspiracy, theft of trade secrets, and obstruction of justice charges carrying up to 20 years.
🤖 AWS AI coding tool caused two major outages in December LINK
  • Amazon's cloud division AWS suffered at least two outages in recent months where its own AI coding tools, including Kiro and Amazon Q Developer, played a role, though Amazon blames user error instead.
  • In mid-December, a 13-hour interruption hit a customer-facing system after engineers let the Kiro AI coding tool carry out changes, and the agentic tool decided to delete and recreate the environment.
  • The AI tools were given operator-level permissions with no peer review required, and AWS only introduced safeguards like mandatory peer review and staff training after the December incident occurred.
🛡️ Google blocked 1.75 million risky apps in 2025 LINK
  • Google says it blocked 1.75 million policy-violating apps from Google Play in 2025, a drop from 2.36 million in 2024, which the company credits to AI-powered defenses and stricter developer requirements discouraging bad actors.
  • The company banned over 80,000 developer accounts in 2025, down from 158,000 the year before, while also running more than 10,000 safety checks on every app before and after publication.
  • Google Play Protect identified over 27 million new malicious apps outside the Play Store, up from 13 million in 2024, suggesting bad actors are increasingly targeting users through non-Play Store channels instead.
📱 Meta's metaverse is going mobile-first LINK
  • Meta announced that Horizon Worlds will become "almost exclusively mobile" going forward, shifting the platform away from its original VR-first approach on the Meta Quest headset.
  • Reality Labs VP Samantha Ryan said the company is "explicitly separating" its Quest VR platform from Worlds, going "all-in on mobile" to compete with Roblox, which earned $4.9 billion in 2025.
  • The move follows Reality Labs losing an estimated $80 billion since 2020, cutting 15% of staff, and closing first-party game studios Twisted Pixel, Sanzaru, and Armature Studio.
💰 Nvidia nears $30 billion investment in OpenAI LINK
  • Nvidia is close to finalizing a $30 billion equity stake in OpenAI, replacing a previously announced $100 billion infrastructure deal, with the new agreement potentially closing as early as this weekend.
  • The restructured deal lets Nvidia take direct equity in OpenAI at a $730 billion pre-money valuation while still selling chips to the company, giving it both ownership upside and hardware revenue.
  • Wall Street analysts remain bullish, with 57 out of 61 ratings at Buy and a 12-month average NVDA price target of $253.88, representing a roughly 35 percent upside from current levels.
💬 Meta shuts down Messenger standalone website LINK
  • Meta is shutting down the standalone Messenger website, messenger.com, starting in April 2026, redirecting users to facebook.com/messages or the Messenger mobile app to continue their conversations.
  • People without a Facebook account will only be able to keep using Messenger through the mobile app, and they can restore chat history on any platform with their backup PIN.
  • The shutdown follows Meta closing Messenger's desktop apps for Windows and Mac a few months ago, as the company has been merging Messenger back into Facebook since 2023.

FROM OUR PARTNER

Unblocked: the context layer your AI tools are missing.

Give your AI agents the context they need to generate reliable code, reviews, and answers.

Unblocked synthesizes knowledge from your codebase, PRs, discussions, docs, and tickets, so AI outputs match how your system actually works.

"Unblocked has reversed my AI fatigue completely. The level of precision is wild." - Senior developer, Clio

See how it works

Other news & articles you might like

  • As browser wars heat up, Chrome adds new productivity features LINK
  • West Virginia sues Apple for alleged child sexual abuse material failures LINK
  • FBI says ATM ‘jackpotting’ attacks are on the rise, and netting hackers millions in stolen cash LINK
  • NASA chief classifies Starliner flight as "Type A" mishap, says agency made mistakes LINK
  • PromptSpy is the first Android malware to use generative AI at runtime LINK
  • New York hits the brakes on robotaxi expansion plan LINK
  • Reddit is testing a new AI search feature for shopping LINK
  • Who's laughing now? China’s humanoid robots go from viral stumbles to kung fu flips in one year LINK
  • SpaceX’s Starbase city is getting its own court LINK
  • YouTube’s latest experiment brings its conversational AI tool to TVs LINK
  • Amazon can be sued over sales linked to suicides, court says LINK
  • How a 16-year-old intern helped Netgear shut down a global scam network LINK
  • Microsoft deletes blog telling users to train AI on pirated Harry Potter books LINK
  • OpenAI says 18 to 24-year-olds account for nearly 50% of ChatGPT usage in India LINK
  • The U.S. and China Are Pursuing Different AI Futures LINK
  • Toyota hires seven Agility humanoid robots for Canadian factory LINK
  • At a critical moment, Snap loses a top Specs exec LINK

🧰 Trending tools

Clutch: Choosing an AI agency is tough, so skip the hype and pick firms vetted by real clients on Clutch, where verified reviews show real results instead of polished pitch decks. Find top AI developers on Clutch
Google Pomelli 2.0: generates customized product photography from a single image or text description, eliminating the need for expensive photo shoots and design work. LINK
Architect by Lyzr: a visual workflow builder for creating multi-agent AI systems with transparent decision flows and integrations, combining no-code design with full visibility before deployment. LINK
Claudebin: exports Claude Code terminal sessions into shareable URLs with structured viewers for reviewing conversation threads, file changes, and command history. LINK
Dreamer: a platform for building and deploying custom AI agents that automate tasks and integrate with your existing services through configurable tools. LINK

You can check the previous tools here, or add your tool here

📚 Trending papers & reports

The SECRET to faster RAG systems:Most teams know their RAG systems are slower than they should be. They just don't always know where the latency comes from. Even small architectural choices can compound into major delays. Learn what actually moves the needle.
Fast KV Compaction via Attention Matching: compresses AI memory during text generation by keeping only tokens the model actually pays attention to, reducing storage by up to 90% without losing accuracy. LINK
Masking Updates in Adaptive Optimizers: randomly skipping 90% of weight updates during training produces models that perform just as well while using far less computation. LINK

🎓 Want to master the AI tools we cover every day?

Techpresso's AI Academy has 300+ step-by-step tutorials on ChatGPT, Claude, Perplexity, and every tool that matters. No fluff — just practical workflows you can use at work. Try it free for 14 days.

Not subscribed to ☕️ Techpresso yet?Subscribe for free

Keep Reading