PLUS: Rivian's robotaxi plans and a new AI bot that can hack better than humans
Good morning
OpenAI has officially launched its GPT-5.2 model series, a new generation designed to excel at professional work. The new models are already showing major performance gains in complex reasoning and real-world tasks, a direct response to a more competitive AI landscape.
With new benchmarks showing the AI can outperform industry professionals across a wide range of occupations, the release raises a critical question: how quickly will this level of capability begin to reshape the job market and the nature of professional work itself?
In today’s Next in AI:
OpenAI's new GPT-5.2 outperforms professionals
Rivian unveils its robotaxi and AI chip plans
TIME names AI's architects Person of the Year
An AI bot that autonomously hacks software
The Next Generation
Next in AI: OpenAI has officially launched GPT-5.2, its most capable model series yet, engineered to excel at professional tasks, coding, and complex reasoning while significantly reducing errors.
Decoded:
The new model series shows notable performance gains in real-world professional work. On the GDPval benchmark, GPT-5.2 Thinking outperforms industry professionals in over 70% of tasks across 44 different occupations, from creating spreadsheets to building presentations.
This leap in capability is powered by cutting-edge infrastructure. OpenAI trained the new models on NVIDIA's latest hardware, including its Hopper and GB200 systems, which are purpose-built for training large-scale AI.
The launch comes at a critical time for the company. Facing intense pressure from competitors like Google's Gemini 3, CEO Sam Altman recently internally declared a 'code red' to accelerate improvements to its core products.
Why It Matters: GPT-5.2 is a direct response to a more competitive AI landscape, aiming to re-establish OpenAI's technical edge with tangible boosts in professional productivity. This move signals a renewed focus on core model competency to deliver more economic value to users and enterprises.
AI Hits the Road
Next in AI: EV maker Rivian held its first Autonomy and AI Day, unveiling its custom-designed AI chip and a next-generation autonomy platform to power its future robotaxi ambitions.
Decoded:
Rivian is building its own silicon with the new RAP1, an in-house processor on a 5nm process that delivers 1600 TOPS of performance for its AI models.
Unlike Tesla's camera-only approach, Rivian confirmed its path to Level 4 autonomy will use a combination of cameras, radar, and LiDAR sensors for a more resilient perception system.
The company announced a new Autonomy+ subscription launching in early 2026 for $49.99 per month, while a software update will soon enable "Universal Hands-Free" driving on millions of miles of roads for current owners.
Why It Matters: Rivian is signaling that automakers must become vertically integrated AI companies to compete in the next decade. By developing its own chip and software stack, the company can accelerate innovation and build a significant challenge to established players like Tesla and Waymo.
AI's Coronation
Next in AI: TIME Magazine has officially named The Architects of AI as its 2025 Person of the Year, highlighting the technology's transformative role this year. The selection cements AI's position as the most consequential force shaping our world.
Decoded:
The recognition highlights figures like Nvidia's Jensen Huang, who predicts AI will grow the global GDP from $100 trillion to $500 trillion.
The award arrives amid significant public anxiety, with polls showing concern and lawmakers proposing bills to ban minors from using chatbots.
This selection follows a historical precedent of honoring transformative technologies, including "The Personal Computer" in 1982 and "You" in 2006 for user-generated content.
Why It Matters: This honor solidifies AI's shift from a speculative technology to a core driver of the global economy and culture. The conversation has now moved from if AI will have an impact to how we navigate the profound changes it brings.
The AI Hacker
Next in AI: A Stanford University experiment showed an AI bot named Artemis can autonomously find and exploit software vulnerabilities, in some cases outperforming professional human hackers according to a new paper.
Decoded:
Artemis operates by scanning a network, identifying potential software bugs, and then crafting its own methods to exploit those vulnerabilities.
Researchers tested the AI in a real-world scenario by unleashing it on the computer network used by Stanford’s own engineering department.
The experiment directly pitted Artemis against professional penetration testers to create a direct performance comparison between the AI and human experts.
Why It Matters: This signals a major advance for AI in cybersecurity, creating powerful new tools for both attackers and defenders. We can expect to see AI-driven vulnerability scanning and automated security testing become industry standard.
The Shortlist
Disney invested $1 billion into OpenAI as part of a landmark deal that will license over 200 of its characters, including those from Marvel and Star Wars, for use in the Sora video generator and ChatGPT Images.
Oracle plunged after its quarterly revenue missed expectations, fueling investor concern that massive capital expenditures on AI infrastructure are not yet generating proportional returns.
Nvidia developed an opt-in software feature that uses telemetry data to verify the country where its data center GPUs are operating, aiming to combat the smuggling of its advanced chips to restricted nations.
SimStudioAI launched a new open-source platform that allows developers to visually build, test, and deploy complex AI agent workflows on a drag-and-drop canvas.