In partnership with
Dear Readers,
The frontier of technology keeps moving faster than anyone can map it. One day it’s a 6,100-qubit quantum leap, the next a world model teaching itself entirely in imagination, and suddenly computer science graduates face a job market reshaped by the very tools they once dreamed of building. The question is no longer whether change is coming, but how ready we are to ride its rhythm.
In today’s issue, you’ll step into that swirl: from quantum processors that flirt with stability once thought impossible, to Dreamer 4’s breakthrough in simulated intelligence, to the uneasy realities of a workforce struggling to find its footing. Add in a glimpse of Sora 2’s app-based debut and Axiom’s AI mathematician promising a discovery flywheel, and the picture sharpens—this is a turning point, both thrilling and unsettling. The stories ahead aren’t just news; they’re signals of the world you’re already living in. Keep reading, and see where the lines connect.
In Today’s Issue:
🎬 Sora 2 just launched as an iOS app, pushing AI video into mainstream social use
✖️ Axiom launched its AI mathematician
🤖 GLM-4.6 is out, upgrading agentic AI with a 200K context window
✨ And more AI goodness…
All the best,
6,100-Qubit Milestone Shocks Physics
A new processor built by Caltech scientists traps 6,100 cesium atoms as qubits—far surpassing prior records for quantum systems. It sustains superpositions for nearly 13 seconds and achieves individual qubit accuracy of 99.98 %, combining both scale and precision. This pushes quantum computing closer to practical, error-corrected machines, though entanglement and processing remain the next big hurdles.
Dreamer 4 Smashes AI Limits
Dreamer 4 introduces a scalable transformer-based world model that trains multi-task agents entirely in imagination, without interacting with real environments. Using shortcut forcing and efficient training, it achieves faster, more accurate predictions and outperforms OpenAI’s VPT agent while requiring 100x less data. This marks a major leap from Dreamer 3, enabling robust, efficient policies across diverse datasets and complex environments.
CS Graduates Struggle Again
Professor Hany Farid warns that computer science graduates today often can’t land meaningful jobs—unlike just a few years ago when offers were plentiful. He argues that AI and market shifts have disrupted the job pipeline, and now students need broader skills and AI fluency to stay competitive.
- Veo 3.1 was (?) planned to launch in ~3 days. Not sure if that's still going ahead in light of Sora 2. Needs to be sufficiently better that it isn't dead on arrival
- Veo 4 launch target is December-ish— #leo 🐾 (#@synthwavedd)
10:12 PM • Sep 30, 2025
The Takeaway
👉 From lab to app — Sora 2 moves AI video into mainstream social use, not just research labs.
👉 Cameos reinvent presence — verified likeness insertion opens new modes of self-expression and collaboration.
👉 Physics + audio realism — more believable outputs expand creative and commercial use cases.
👉 Platform stakes — shows how AI companies may compete directly with TikTok-style ecosystems.
A new chapter in digital storytelling just opened: Sora 2 isn’t only OpenAI’s most advanced video-generation model — it now arrives as a dedicated iOS app. Think TikTok, but every clip is AI-native. Users can generate cinematic scenes, remix others’ creations, and even drop themselves or friends into videos via “cameos,” all verified with a one-time likeness capture.
Technically, Sora 2 is a major leap. It syncs audio with visuals, respects physics (a basketball bounces instead of teleporting), and follows multi-shot instructions with consistency. That makes outputs both more controllable and more believable. But the app format changes the game: it transforms world simulation from a research milestone into a social, co-creative experience where entertainment, creativity, and community intersect.
Why is this so important for AI? Because Sora 2 hints at platforms where people don’t just consume feeds but co-build synthetic realities. It’s a testbed for the future of simulation — wrapped in a playful, viral format.
Why it matters: By merging AI breakthroughs with a social app, OpenAI is stress-testing how society interacts with synthetic media at scale. The implications stretch from creativity and communication to ethics and digital identity.
Sources:
The best HR advice comes from people who’ve been in the trenches.
That’s what this newsletter delivers.
I Hate it Here is your insider’s guide to surviving and thriving in HR, from someone who’s been there. It’s not about theory or buzzwords — it’s about practical, real-world advice for navigating everything from tricky managers to messy policies.
Every newsletter is written by Hebba Youssef — a Chief People Officer who’s seen it all and is here to share what actually works (and what doesn’t). We’re talking real talk, real strategies, and real support — all with a side of humor to keep you sane.
Because HR shouldn’t feel like a thankless job. And you shouldn’t feel alone in it.
The SWE Bench is expected to be 90% saturated as early as January 2026.
AI Mathematician Sparks Discovery Flywheel
Axiom has officially launched, introducing its mission to build a self-improving superintelligent reasoner, beginning with an AI mathematician. The project emphasizes that mathematics has historically served as the foundation for breakthroughs across science, technology, and human understanding, and AI now accelerates this process dramatically. Carina Hong highlights a “renaissance” powered by three pillars: AI’s capacity to generate vast ideas, programming languages’ ability to translate abstraction into reality, and mathematics as the unifying framework.
Today, I am launching @axiommathai
At Axiom, we are building a self-improving superintelligent reasoner, starting with an AI mathematician.
— #Carina Hong (#@CarinaLHong)
2:23 PM • Sep 30, 2025
Together, these elements create a discovery flywheel where each advancement compounds the next. The team behind Axiom includes pioneers in synchronous training, deep learning for code, and applying Transformers to solve long-standing math problems. Their vision is to make an AI mathematician accessible to everyone, catalyzing research and innovation across fields. The call to action invites others to join Axiom’s mission and shape this new era of accelerated discovery.
GLM-4.6: Next-Gen Agent Engine
GLM-4.6 upgrades GLM-4.5 with a 200K token context window, improved reasoning, enhanced coding benchmarks, better tool-use, and writing refined for human preferences. It competes with Claude Sonnet 4 on multiple public benchmarks and powers agentic applications (Claude Code, Roo Code, etc.). Efficiency gains (≈30% lower token consumption vs. 4.5) make it more practical for real-world deployment.
Launch, grow, & monetize your newsletter with Beehiiv
It's the simplest platform to start, yet powerful enough to scale with your ambitions. Effortlessly build your audience and transform your content into a thriving business.
