Tuesday 08 July 2025 | Join Free | Upgrade

Together with Sponsor

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🚶‍♂️ Meta just hired Apple’s head of foundation models

🔒 OpenAI activates military-grade security to protect its AI models

📈 Threads is nearing X’s daily app users

📱 Apple tones down Liquid Glass after user complaints

💰 OpenAI fights Meta with $4.4 billion stock pay

🙏 Cursor apologizes for unclear pricing changes

🎁 + 12 other news you might like

🔮 + 6 handpicked research papers and tools

FROM OUR PARTNER


Industry disruptors don’t compete—they change the rules entirely

Uber didn’t wait for New York to approve ride-hailing apps. They got drivers on the streets and users on the app.

And from a $1.25M seed round in 2010, they turned into a $195 billion behemoth.

That’s why investors are now eyeing Mode Mobile.

They’re changing the rules behind who earns from data, allowing users to earn real money from their screen time.

Here’s what that looks like:

And accredited investors can still invest at $0.30/share with up to a 120% bonus.

🚶‍♂️ Meta just hired Apple’s head of foundation models LINK
  • Ruoming Pang, the engineering manager for the core models team behind “Apple Intelligence,” has departed the company to join competitor Meta in a multi-million-dollar deal.
  • The exit underscores turmoil inside Apple's AI division, where morale was hurt by discussions to use outside technology from other companies to power a future Siri.
  • This poach reveals significant technical vulnerabilities, as Apple's advanced Siri features are delayed until 2026 for a complete “V2” architectural rebuild from the ground up.
🔒 OpenAI activates military-grade security to protect its AI models LINK
  • OpenAI implemented a "deny-by-default" internet policy and uses information "tenting" to restrict employee access and stop leaks of its foundational model technologies.
  • The company installed biometric fingerprint scans and hired a former Palantir CISO and a retired U.S. Army General to oversee its cyber and data defense.
  • These security upgrades follow allegations that Chinese rival DeepSeek used a technique known as “distillation” to copy OpenAI’s models and build its own system.
📈 Threads is nearing X’s daily app users LINK
  • Threads' worldwide daily active users on mobile apps reached 115.1 million, getting closer to X's declining figure of 132 million daily actives.
  • However, X maintains a clear advantage on the web, with 145.8 million average daily web visits worldwide compared to just 6.9 million for Threads.
  • In the U.S. market specifically, X still leads on mobile with 22.9 million daily active users, ahead of the 15.3 million reported for Threads.
📱 Apple tones down Liquid Glass after user complaints LINK
  • Apple's third developer beta for iOS 26 tones down the Liquid Glass effect, making the translucent, glass-like elements in navigation bars less pronounced.
  • Liquid Glass is a dynamic material that adds depth to the interface by reflecting its surroundings on app icons, widgets, controls, and other components.
  • Following this change, some testers expressed disappointment on social media, feeling the reduced visual flair takes away the futuristic feel of previous beta releases.
💰 OpenAI fights Meta with $4.4 billion stock pay LINK
  • OpenAI is defending against Meta by increasing its stock-based compensation to $4.4 billion, a figure that represents 119 percent of its revenue from last year.
  • Meta poached eight researchers with reported nine-figure offers after its own Llama 4 “Behemoth” model failed performance benchmarks, prompting a period of internal panic.
  • The rival formalized its raid by creating Meta Superintelligence Labs, forcing OpenAI leadership to promise in a leaked memo they were “recalibrating comp” to retain talent.
🙏 Cursor apologizes for unclear pricing changes LINK
  • The company behind the coding tool Cursor apologized for poorly communicating a pricing change that switched its Pro plan from 500 fast responses to a credit system.
  • Pro users quickly depleted their new $20 worth of usage, especially with expensive Claude models, resulting in some being unexpectedly charged for additional costs they did not anticipate.
  • Anysphere is now refunding affected subscribers, explaining the change was necessary to pass along the high cost of running the latest and more expensive AI models.

FROM OUR PARTNER

Go Beyond Vibe Coding, Let AI Deploy Live Apps

You don’t need to be a full-stack engineer to ship a full product. 

VibeOps turns your vibes into a live app:

  • Configures databases

  • Sets up backend

  • Adds functioning authentication

Turn prototypes into full-fledged live products in minutes. Check out our demo and join our closed beta waitlist now!

Join the waitlist now

Other news & articles you might like

  • Apple and Masimo back in court over Apple Watch import ban appeal LINK
  • Robinhood CEO downplays OpenAI concerns on tokenized stock structure LINK
  • Waymo robotaxis are heading to Philadelphia and NYC LINK
  • Unless users take action, Android will let Gemini access third-party apps LINK
  • Sakana AI's new algorithm lets large language models work together to solve complex problems LINK
  • Epic reaches mystery settlement with Samsung days before new Galaxy phones LINK
  • xAI updated Grok to be more ‘politically incorrect’ LINK
  • This New AI Tool Wants to Work With Filmmakers—Not Replace Them LINK
  • Samsung expects second-quarter profits to more than halve as it struggles to capture AI demand LINK
  • Microsoft: Edge Now Loads Its UI 40% Faster Than Before LINK
  • Apple’s newest AI study unlocks street navigation for blind users LINK
  • China’s robot dog breaks Boston Dynamics’ speed record, runs 100m in 13.17 seconds LINK

Trending research and tools

The Lifespan of our Universe: current theories suggest the universe will continue to expand forever until it becomes a cold, dark, and empty place after all stars have died and matter has decayed. LINK
Query Agnostic Adversarial Triggers for Reasoning Models: researchers found it is possible to create short, universal phrases that can be attached to any prompt to consistently make AI models give wrong answers. LINK
Energy-Based Transformers Are Scalable Learners and Thinkers: the paper introduces a new type of AI model that gets better with more data and can solve complex, multi-step problems. LINK
Because We Have LLMs, We Can and Should Pursue Agentic Interpretability: the paper proposes we should try to understand how large language models work by treating them as agents with their own goals and beliefs. LINK
AsyncFlow: An Asynchronous Streaming RL Framework for LLM Post-Training: the paper presents a system that uses an asynchronous, streaming approach to more efficiently fine-tune large language models with reinforcement learning. LINK
DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Gen: this paper presents a new model that improves code generation by using a masked diffusion technique to fill in missing sections of code. LINK
Disclosures

Mode Mobile recently received their ticker reservation with Nasdaq ($MODE), indicating an intent to IPO in the next 24 months. An intent to IPO is no guarantee that an actual IPO will occur.

The Deloitte rankings are based on submitted applications and public company database research, with winners selected based on their fiscal-year revenue growth percentage over a three-year period.

Uber’s valuation listed as the total market capitalization calculated as of July 1st, 2025 market close. Data taken from Yahoo Finance.

This offer is only open to accredited investors.

Want to get the latest news differently?

Not subscribed to ☕️ Techpresso yet? Subscribe for free