☕️ Meta tests paid subscriptions on Instagram

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

👀 Meta tests paid subscriptions on Instagram

🐀 Software tool used by millions compromised in hack

👓 Meta launches prescription AI glasses at $499

🛰️ Another Starlink satellite has inexplicably exploded

📧 Google now lets US users change their Gmail address

Plus: 🎁 17 other news you might like, 🧰 6 tools, and 📚 3 papers.

👀 Meta tests paid subscriptions on Instagram LINK

Meta is testing a paid subscription called Instagram Plus in a few countries, giving everyday users access to exclusive features for a small monthly fee.

Subscribers can view Stories without the poster knowing, see how many people rewatched their Stories, create unlimited audience lists, and extend or spotlight their Stories.

Pricing varies by country — roughly $1 to $2 per month in Mexico, Japan, and the Philippines — and the subscription is separate from Meta Verified, which targets creators and businesses.

🐀 Software tool used by millions compromised in hack LINK

An attacker hijacked the npm account of a lead Axios maintainer and published two malicious versions of the popular JavaScript HTTP client library, which is downloaded about 100 million times per week.

The poisoned releases added a fake dependency called plain-crypto-js that silently installed a cross-platform remote access trojan on developer machines running macOS, Windows, and Linux within seconds of installation.

The malicious versions were live for roughly two to three hours before npm removed them, and security firms recommend treating any system that ran the package as fully compromised.

👓 Meta launches prescription AI glasses at $499 LINK

Meta and EssilorLuxottica launched two new Ray-Ban Meta smart glasses styles, the Blayzer Optics and Scriber Optics, starting at $499 and designed to work with prescription lenses including progressive and transition options.

Both frames feature slimmer designs, swappable nosepads, and adjustable temple tips for a better fit, and they are available for pre-order now with sales starting April 14 at retailers like LensCrafters.

Meta is also rolling out new software features for all Ray-Ban Meta glasses, including Japanese, Mandarin, and Arabic translation support, food and nutrition tracking, and AI-powered message thread summaries.

🛰️ Another Starlink satellite has inexplicably exploded LINK

SpaceX has lost contact with another Starlink satellite after an unexplained "anomaly," and tracking company Leo Labs detected tens of debris objects near Starlink 34343 shortly after the event.

SpaceX says the event poses no risk to the Space Station, its crew, or NASA's Artemis II launch, and the satellite fragments are expected to burn up in the atmosphere within weeks.

The explosion happened at about 560km above Earth in low Earth orbit, where over 24,000 objects — including around 10,000 Starlink satellites — are currently being tracked in an increasingly crowded zone.

📧 Google now lets US users change their Gmail address LINK

Google is now rolling out a feature that lets users in the U.S. change their Gmail address without creating a new account or losing access to their existing data.

Users can only change their username once every 12 months, and their old email address will be preserved as an alternate address that still works for signing in.

The feature is rolling out gradually, so not everyone will see the "Change Google Account email" button in their Personal info settings right away, according to Google's support page.

Other news you might like

• Newsom Signs AI Safety Order for California State Contracts, Defying TrumpLINK

• Amazon Leo ramps Starlink battle with Delta Wi-Fi dealLINK

• Apple Intelligence shows up in China early, then gets shut down fastLINK

• 150 million users later, Roblox competitor Rec Room is shutting downLINK

• As more Americans adopt AI tools, fewer say they can trust the resultsLINK

• Apple steps up crackdown on vibe coding apps, pulls ‘Anything’ from the App StoreLINK

• Delve whistleblower strikes again, with alleged receipts about ‘fake compliance’LINK

• GitHub backs down, kills Copilot pull-request ads after backlashLINK

• After 16 years and $8 billion, the military's new GPS software still doesn't workLINK

• It's not just a RAM crisis — Panasonic says data center batteries are also selling out months in advanceLINK

• Anthropic accidentally leaked Claude Code source code via a map file in npm registry, revealing hidden “Capybara” models and AI petLINK

• OpenAI Ships a Codex Plugin for Claude Code, Putting Its Agent Inside a Rival's ToolLINK

• A man used AI to call 3,000 Irish bartenders to track the cost of Guinness. Now pubs are lowering their prices to competeLINK

• Robots just installed 100 MW of solar panels in the desertLINK

• Between JobsLINK

🧰 Trending tools

Jupid: a financial data layer that connects your bank, categorizes transactions for IRS Schedule C with 96% accuracy, and remembers vendor relationships across sessions.LINK

30u30.fyi: tracks which Forbes 30 Under 30 alumni have been charged with crimes or involved in fraudulent activities.LINK

Pixero AI: automates Meta ad campaigns from strategy to deployment by scraping your URL, generating creative assets, and managing campaigns in under 10 minutes.LINK

Computer Use in Claude Code: enables Claude AI to control a computer interface directly, executing tasks like clicking, typing, and navigating applications through visual screen analysis.LINK

Solvea: an AI receptionist platform that handles customer support, sales, and appointment scheduling across phone, chat, and email with contextual memory and no coding required.LINK

Perplexity API Platform: unified API providing multi-model access, real-time web search across 200B+ URLs, and embeddings through a single integration with direct provider pricing.LINK

📚 Trending papers & reports

AI coding agents working in parallel complete tasks 2.3 times faster than solo agents by splitting work into independent subtasks, with accuracy remaining above 85% on software engineering benchmarks.LINK

Removing "to be" verbs from an LLM's vocabulary improved reasoning accuracy by forcing the model to use more concrete, action-oriented language in its chain of thought process.LINK

ARC-AGI-3 benchmark introduces 100 new visual reasoning puzzles where current AI systems score under 5%, revealing that frontier models still struggle with novel abstract pattern recognition tasks.LINK

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

See you tomorrow for a new dose of ☕️ Techpresso!