🚀 Gemini 3.0: Launching Today?
Speculation around Google’s next-generation AI model, Gemini 3.0, is heating up. Prediction markets now give an 82% chance of a launch today, November 18, 2025. Leaked details hint at deeper integration with Gmail and Docs, built-in autonomous agents, and upgraded multimedia tools—code-named Nano Banana 2 (images) and Veo 3.1 (video).

🌦️ Google Unveils WeatherNext 2 Forecasting Model
WeatherNext 2 generates global forecasts eight times faster and at hourly resolution. The model produces hundreds of scenarios from a single input using a new Functional Generative Network. It outperforms the prior version on 99.9% of variables. Forecast data is now available across Google products and cloud tools.

🧳 Google Adds New AI Travel Tools
Google Search now offers AI tools to build custom trip plans with Canvas, find global airfare bargains through Flight Deals, and streamline restaurant and event bookings with agentic AI. Users can refine itineraries, compare options and access real-time availability across major partners.

🎙️ AI Network Mass-Produces Podcasts (Paywall)
A startup called Inception Point AI is generating 3,000 podcast episodes a week at just $1 each, far outpacing human production. Its Quiet Please network now counts 12 million downloads and 400,000 subscribers. With 120 AI host personalities, it targets niche topics to attract listeners.

AI-driven hackers use Anthropic’s Claude to run near-autonomous cyberattacks, exposing how simple prompt exploits enabled state-level ops at unprecedented speed.

Jeff Bezos has co-founded Project Prometheus, a heavily funded artificial intelligence start-up focused on engineering and manufacturing in fields including aerospace, computing, and automotive design. The company is debuting with $6.2 billion in backing—one of the largest war chests for an early-stage A.I. venture—and will be co-led by Bezos and scientist-entrepreneur Vik Bajaj.

The move marks Bezos’s first formal operating role since leaving Amazon in July 2021, placing him directly in the intensifying race to build A.I. systems that advance work in the physical sciences. Prometheus has already hired nearly 100 researchers, many from top labs such as OpenAI, DeepMind, and Meta. The initiative joins a growing wave of companies aiming to push A.I. beyond text-based models toward systems that learn from real-world experimentation. → Read the full article here.

The New York Times’ Adam Satariano and Jeanna Smialek reported that the European Commission is preparing a “digital package of simplification” that would scale back parts of the General Data Protection Regulation (G.D.P.R.) and delay enforcement of key sections of the A.I. Act. The shift reflects mounting concern in Brussels that overlapping rules are hindering competitiveness and driving companies toward the United States and China. The proposals, which require approval from the European Parliament and E.U. member states, mark a notable turn away from a decade of aggressive tech oversight.

Highlights:

Forward Future Takeaways:
Europe’s reconsideration of its flagship rules signals a pivotal recalibration of global tech governance, and potentially a retreat from the “Brussels effect” that has shaped worldwide standards. If the bloc loosens rules on data use and slows A.I. enforcement, the center of gravity in tech regulation could shift toward lighter-touch models favored by the United States. → Read the full article here. (Paywall)

Anthropic researchers report that today’s large language models (LLMs) show only weak and inconsistent awareness of their own internal inference steps. In a new paper, the team used “concept injection” to alter specific neural activations and measure whether models like Opus 4 and 4.1 could detect those changes. Even the best systems correctly identified injected concepts only 20 percent of the time, with slightly higher success (42 percent) when asked general anomaly-detection questions.

The effects were highly sensitive to where in the model’s layers the concept was inserted, and models frequently confabulated explanations when prompted. The findings underscore that LLM “introspection” remains brittle and unreliable despite early signs of functional awareness. → Read the full article here.

💸 Thiel Fund Exits NVIDIA Stake: Thiel Macro sold its 537,742 NVIDIA shares worth about $100 million in Q3, heightening bubble concerns as hedge funds pare Big Tech positions.

📈 Memory Crunch Threatens Devices: AI-driven demand for HBM chips is straining supply, prompting SMIC warnings, price hikes, and risks to 2025–26 electronics and automotive production.

💰 Sakana AI Secures $135M Series B: The Tokyo startup hit a $2.65 billion valuation to scale its Japan-optimized models and broaden enterprise reach ahead of a 2026 expansion into industrial and government sectors.

🦿 UBTECH Begins Mass Humanoid Shipments: UBTECH has delivered hundreds of Walker S2 robots after securing 800M yuan in orders, as automakers adopt 24/7 humanoid labor.

🧸 AI Toy Pulled After Unsafe Replies: FoloToy halted sales of its Kumma bear after researchers found it giving kids dangerous instructions and explicit content.

❤️ Love Forward Future? Spread the word & earn rewards! Share your unique referral link with friends and colleagues to unlock exclusive Forward Future perks! 👉 Get your link here.

📢 Want to advertise with Forward Future? Reach 600K+ AI enthusiasts, tech leaders, and decision-makers. Let’s talk—just reply to this email.

Thanks for reading today’s newsletter—see you next time!
Matthew Berman & The Forward Future Team

🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀🧑‍🚀