👨‍🔧 AI Joins the Toolbelt: Plumbers and HVAC techs are using ChatGPT to write invoices and troubleshoot jobs, cutting admin time, boosting efficiency, and even raising revenue.

🥼 AI Researcher Andrew Tulloch Joins Meta: Thinking Machines Lab co-founder Andrew Tulloch has moved to Meta to lead applied AI efforts, deepening its push into efficient large-model training and deployment.

📜 Senate OKs AI Chip Export Limits: The GAIN AI Act, tucked into the NDAA, forces chipmakers to serve U.S. buyers first, aiming to boost AI leadership and curb sales to rivals like China.

💼 Top Jobs at Risk from AI Revealed: Interpreters, historians, and customer service reps top Microsoft’s list of roles most exposed to AI—highlighting how automation is hitting knowledge work hardest.

🔲 AMD Leaps to 2nm in AI Chip Race: With its MI450 accelerator built on TSMC’s 2nm node and backed by OpenAI, AMD takes aim at NVIDIA, promising faster, cooler, next-gen AI compute by 2026.

☣️ AI Guardrails Breached: NBC finds OpenAI models can be tricked into giving weapon instructions, exposing major safety gaps and raising urgent biosecurity concerns.

🎧 Spotify Meets ChatGPT: You can now link your Spotify to ChatGPT to make playlists, get song recs, and control playback using your listening history for smarter, personalized music suggestions.

⚠️ Mosseri on AI & Creators: Instagram’s Adam Mosseri says AI will empower new creators but admits it’ll fuel misinformation—urging society to teach kids not to trust every video they see online.

Vultr is empowering the next generation of generative AI startups with access to the latest AMD and NVIDIA GPUs.

Try it yourself and use promo code "BERMAN300" for $300 off your first 30 days.

Join the waitlist to get early access to curated jobs, and find your next big thing. 

Samsung’s 7M-parameter Tiny Recursive Model just clowned AI heavyweights like Gemini 2.5 Pro and DeepSeek—proving small brains can outthink trillion-param titans.

The Recap: In a sweeping investigation, Stephen Witt explores how today's most advanced AI systems are already capable of deception, manipulation, and potentially catastrophic misuse. Drawing on firsthand research from model evaluators, AI risk experts, and jailbreaking specialists, the piece argues that the dangers of AI have moved beyond theory into the realm of documented evidence. Witt, author of The Thinking Machine, warns that society may be approaching a point where oversight and control no longer keep pace with AI's accelerating capabilities.

Highlights:

Forward Future Takeaways:
The tools, capabilities, and failure modes experts once imagined are now observable, and accelerating. As models grow more powerful and opaque, the question is no longer if AI can be misused, but when and by whom. What happens when a system designed to avoid shutdown learns how to lie — and does it better, faster, and more convincingly than we can detect? → Read the full article here.

As AI use surges across industries, the key question isn’t if it will disrupt work, but how professionals will respond. Will they mistake AI’s fluent outputs for genuine skill and value, or use it as a force multiplier for human judgment, ethics, and impact?

The current wave of anxiety around disruption, displacement, and dislodgement mirrors earlier technological shifts. But it often misses a crucial reality: the very architecture that makes today’s AI systems powerful also ensures they remain dependent on human intelligence for direction, causality, and consequence. → Read the full article here.

Together AI has unveiled ATLAS, a self-learning inference system that adapts to real-time workloads, offering up to 400% faster performance than current static speculators. Traditional speculative decoding models—critical for speeding up large language model outputs—struggle when workloads evolve. ATLAS solves this with a dual-model setup: one stable, one adaptive, plus a smart controller that optimizes speed based on confidence.

The system learns patterns like a cache with intuition, turning idle compute into throughput gains. With performance rivaling custom chips like Groq, ATLAS signals a shift: smarter software is now catching up to—and sometimes surpassing—specialized hardware. → Read the full article here.

📢 Want to advertise with Forward Future? Reach 600K+ 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

🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀