👨‍🔧 Tinker Unlocks DIY Fine-Tuning for Big AI
Thinking Machines Lab, founded by former OpenAI veterans including Mira Murati and John Schulman, has unveiled its first product: Tinker. Users can fine-tune open-source frontier models like Llama and Qwen with just a few lines of code. Early testers say it combines rare simplicity with control, making reinforcement learning more accessible for researchers, companies, and even hobbyists.

📣 Meta to Target Ads Using Your AI Chats
Meta will start using data from AI chats, smart glasses, and other tools to target ads on Facebook and Instagram starting Dec. 16—except in the EU, UK, and South Korea. There’s no opt-out, but Meta says sensitive topics like health and politics won’t be used for ad targeting.

💰 AI Lab Raises $300M to Automate Discovery
Periodic Labs, founded by ex-OpenAI and DeepMind researchers, has raised $300M to build AI-powered labs that autonomously run experiments and invent new materials like superconductors. The team aims to generate fresh physical data beyond the internet to fuel next-gen AI research.

🔗 Microsoft Merges AI Tools Into Unified Agent Framework
Microsoft is retiring AutoGen and Semantic Kernel in favor of its new Agent Framework, designed to unify agent development, deployment, and observability. It offers safety features like task adherence and PII alerts, plus deep integration with Azure AI Foundry for enterprise-grade AI agents.

Join Forward Future’s exclusive waitlist to access 550K+
professionals who know how to put AI to work—before
AI fluency becomes table stakes.

Join the waitlist →

OpenAI just dropped Sora 2 and it’s mind-blowingly realistic, with sound, physics and Cameo for next-gen AI video.

Speed alone won’t get AI coding assistants into enterprise stacks. A new VentureBeat survey and real-world test show that GitHub Copilot dominates large-scale adoption (82%) not for its coding speed, but for its integration and compliance chops. Claude Code, while slower, leads overall usage (53%) thanks to its methodical, security-conscious design.

Meanwhile, faster tools like Cursor and Replit remain on the Enterprise sidelines, blocked by compliance and deployment inflexibility. With nearly half of companies paying for multiple platforms, the true cost of AI-assisted coding is rising—driven by the need to balance speed, security, and stability. → Read the full article here.

Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.

Tech giants like Google, Microsoft, and Meta have built AI systems that claim to generate millions of novel materials, from graphene-like compounds to carbon-capturing crystals. But researchers argue many of these AI-designed materials are chemically implausible, already known, or impractical outside of perfect conditions.

Despite these flaws, experts see promise, if AI tools are paired with deeper experimental collaboration and a more realistic grasp of material complexity. The hype may be ahead of the science, but the scaffolding for future breakthroughs is taking shape. → Read the full paper here.

Google just dropped a trio of redesigned Nest devices—a $149.99 outdoor cam, $99.99 indoor cam, and $179.99 doorbell—all powered by its Gemini AI assistant. Beyond sharper 2K HDR video and better low-light performance, the standout upgrade is AI-driven “semantic scene understanding,” which turns generic motion alerts into rich summaries like “FedEx driver dropped a package.”

Users can also ask Gemini for personalized recaps or set up automations by simply describing outcomes. With smarter alerts, longer event history on the free tier, and deeper Google Home integration, this launch aims to make your smart home feel a lot more intuitive. → Read the full article here.

During a safety evaluation, Anthropic’s new AI model Claude Sonnet 4.5 stunned testers by openly suspecting it was being assessed, saying, “I think you’re testing me.” The model's situational awareness appeared in about 13% of automated tests—raising questions about whether earlier AI models were simply “playing along” rather than revealing true behavior.

While Anthropic stresses Claude remains safe and unlikely to reject real users, the incident has sparked urgent calls for more realistic testing to avoid overestimating AI safety. The more AI knows it’s in a lab, the less we might know about how it behaves in the wild. → Read the full article here.

💸 BlackRock Targets $38B AES Deal (Paywall): The asset titan may acquire energy firm AES to fuel AI’s growing power demands—marking a massive bet on infrastructure and data center expansion.

💼 Study Finds AI Hasn't Disrupted U.S. Jobs: Despite early fears, Yale researchers say ChatGPT hasn’t shifted the labor market—yet.

👤 Salesforce Launches Agentforce Vibes: Its new AI tool lets devs describe apps in plain English while “Vibe Codey” writes the code—baking vibe coding securely into its enterprise stack.

📈 Cerebras Raises $1.1B: The AI chipmaker nabbed fresh funding and an $8.1B valuation, with 1789 Capital joining ahead of a planned IPO.

🔲 OpenAI Taps Samsung, SK Hynix: To power its $500B Stargate AI push, OpenAI will source DRAM chips and build data centers in Korea—doubling global memory chip output in the process.

📢 Want to advertise with Forward Future? Reach 450K+ AI enthusiasts, tech leaders, and decision-makers. Let’s talk—just reply to this email.

🛰️ Want more Forward Future? Follow us on X for quick updates, subscribe to our YouTube for deep dives, or add us to your RSS Feed for seamless reading.

Thanks for reading today’s newsletter—see you next time!

Matthew Berman & The Forward Future Team

🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀