PRODUCTS

OpenAI’s GPT-5.2 is really good at spreadsheets

OpenAI rolled out an upgraded version of ChatGPT following a turbulent week in which it declared a “code red” in light of recent advances by Google Gemini.

Although some OpenAI employees reportedly lobbied to delay the model’s release, the new version, GPT-5.2, outperformed GPT-5.1 on several work, coding and general benchmarks. GPT-5.2 debuted at #2 on LMArena’s web development leaderboard, behind Claude Opus 4.5. Code red will remain in place until a wider model release happens in January, the WSJ reported, but for now, there may be a little less heat on OpenAI’s garlic.

OpenAI has taken some flak lately for cutting behemoth compute deals and CEO Sam Altman’s perceived lack of focus on ChatGPT. With that in mind, much of OpenAI’s messaging centers on the economic value of its latest model.

OpenAI’s press release introduces GPT-5.2 as “the most capable model series yet for professional knowledge work” and highlights its capabilities in white-collar work. OpenAI vice president Nick Turley and president Greg Brockman both piped in on X to praise the new model’s spreadsheet skills. 

GPT-5.2’s release, along with OpenAI’s massive Disney deal, spells good news for the company, but it may not be champagne time in the C-suite just yet.

The family of an 83-year-old woman who was murdered by her son after he had ChatGPT-enabled hallucinations sued OpenAI for wrongful death. This latest legal action follows seven suits that were filed in November, making wrongful death, assisted suicide and involuntary manslaughter claims against the ChatGPT maker.

OpenAI said GPT-5.2 performed better in a mental health context than its prior model and noted that it would roll out an age prediction model to limit exposure to sensitive content for users under 18.

After Anthropic CEO Dario Amodei chided Sam Altman for liking to “YOLO things” and a report emerged that Altman considered getting into the space business last year, it felt like OpenAI needed to assure investors and the public that it is focused on delivering an economically valuable chatbot product. GPT-5.2 is that. Still, the new wrongful death lawsuit filing indicates that things won’t get easier from here. OpenAI may be in for safety-related backlash from public officials and the public.

TOGETHER WITH TEMPORAL

A look inside the AI stack of 150+ leading organizations

Your AI demo shines. But can it scale?

We surveyed engineers to expose the hidden gaps in today’s AI tech stacks and how to build ones that last.

Download your free copy of “The 2025 Production AI Stack Report” right here.

CULTURE

Disney chooses sides in the AI war

As Hollywood debates where AI fits in, Disney has chosen which horse it wants to back. 

On Thursday, the entertainment giant announced a licensing deal with OpenAI, allowing its video generation platform Sora to generate shorts based on more than 200 characters from its canon, including from Disney, Marvel, Pixar and Star Wars IP. Certain Sora shorts will be available to stream on Disney+ as part of the agreement.

Also, as part of the deal, Disney will become a “major customer” of OpenAI, integrating its tools to build products and experiences and deploying ChatGPT to its employees. Disney will also make a $1 billion equity investment in the AI firm. 

Iger told CNBC on Thursday that the deal “does not in any way represent a threat to the creators at all” as it excludes access to likenesses and character voices.

In placing its bets on OpenAI, Disney immediately made an enemy of another AI giant: Google. On Wednesday night, the company sent Google a cease-and-desist letter, accusing the company of AI-based copyright infringement on a “massive scale.” Disney has sent similar cease-and-desist letters to Meta and Character.AI and has filed joint litigation with NBCUniversal and Warner Bros. Discover against Midjourney and Minimax. 

“Disney isn't playing around. By partnering with OpenAI, they're picking someone they can work with while setting the standard for how Hollywood deals with AI,” Darrell White, partner at law firm Kimura London & White, told The Deep View. “The takeaway is simple: AI is coming to entertainment, but only if the studios call the shots.”

Disney picking its winner is another signal that the industry is choosing sides as Google rapidly catches up to OpenAI. Reports emerged last week that OpenAI CEO Sam Altman signaled a “code red” alert to employees, demanding all hands on deck to improve its flagship chatbot. Altman also warned staff in late November that Google’s success with Gemini would cause economic headwinds.

Getting the endorsement of one of the biggest players in entertainment (not to mention debuting GPT-5.2) gives OpenAI another leg to stand on, as Google’s gains have made it sweat.

Though the AI model battle has made its way to Hollywood, there’s still significant pushback against the tech from artists and the people representing them. When Sora was first introduced in October, the model was widely criticized by agencies, with Hollywood’s three talent powerhouses – Creative Artists Agency, United Talent Agency and William Morris Endeavor – opting their clients out of the platform. While some facets of the industry have started warming up to the tech, such as major music labels forging deals with AI platforms like Suno and Udio, opinions are mixed, and tension is high as AI creates a potentially existential threat to something that’s innately human: The creation of art.

TOGETHER WITH AUGMENT CODE

Catch critical issues without the noise

Augment Code Review’s retrieval engine pulls the exact set of files and relationships necessary for the model to reason about cross-file logic, API contracts, concurrency behavior, and subtle invariants, translating directly into higher recall (55%) without sacrificing precision (65%), and putting Augment 10 pts ahead of the competition on combined recall and precision.

Augment Code Review helps your team succeed with: 

• a superior context engine

• purpose-built code review 

• a reasoning-driven agent loop

Check out Augment Code Review and catch real bugs without spamming your PRs!

WORKFORCE

Companies turn bearish on AI adoption, based on fear

At the AI Summit in New York City, technology leaders from AstraZeneca, Nvidia, the United Nations, and smaller firms debated how deeply AI should shape organizational decision-making, from internal corporate workflows to humanitarian operations.

Their consensus: AI can help organizations move faster and work smarter, but fear of risk is slowing adoption.

“Very few clients want the whole process automated,” Graham Wilkinson, Chief Innovation Officer at Acxiom, a data services firm powering AI-driven marketing, said during the December 10 panel.

While 68% of business leaders view AI as an opportunity, the share who see it as a risk has doubled since 2024, according to a 2025 survey from Gallagher, a risk management firm. A separate EY study from this year found that companies lost an average of 4.4 million dollars on early AI deployments, largely because of compliance missteps, faulty outputs, and security problems. 

Risks like these, Wilkinson said, have left many companies “paralyzed” at the start of their AI journeys. “No one wants to be the mistake that the news blows up,” he said.

The public sector faces similar hesitation. Anusha Dandapani, chief of data and AI at the United Nations International Computing Centre (UNICC), said responsible AI in humanitarian work requires strong design, accountability, and models that work across languages and cultures. Many regions lack AI-ready data, she added, which often derails pilots.

AstraZeneca chief architect Wayne Filin-Matthews said the company is embedding agentic AI into its workflows with caution. Across research and development, finance, and IT, the pharmaceutical giant uses a “risk-based approach” to deployment, prioritizing transparency, security, and regulatory compliance.

AstraZeneca’s biggest challenge is ensuring that the “semantic kernel,” the layer connecting AI models to enterprise code, does not compromise security. To address this, the company launched “Project Streamline” to cut delays from privacy checks and help technical and business teams work together more smoothly.

By the end of the panel, speakers agreed that getting AI right is not just about speed but managing risk with precision. As Nvidia’s OEM Alliance Manager Matt Hitt put it: “It's important to start small, measure big, and evaluate” before scaling.

LINKS

• Google appointed a new “chief technologist for AI infrastructure”

• Rivian introduced its own chip for its autonomous driving efforts, replacing Nvidia chips.

• Health and fitness searches are the top use for Microsoft Copilot.

• WeChat operator Tencent is finally getting into the AI game. 

• Stanford researchers built an AI bot that was better at hacking than 9 out of 10 “network penetration testers.”

• AI voice company Eleven Labs is partnering with Meta. 

• A bill focused on improving AI talent in federal agencies is set to be introduced in Congress.

• Disco: Gemini-powered app that turns open tabs into custom applications

• Cursor visual editor: Limited-code editor for designers

• Google MCP support: Google rolled out fully-managed, remote MCP servers to enhance AI agent development

• Nano Banana Elements: Create sets of styles, objects, or characters for use in Nano Banana prompts

• Stitch Prototypes: Google Labs no-code tool for designers

The Forward Deployed Engineer isn’t another hire—it’s a new way to build.

Forward Deployed Engineers embed where decisions happen, translating business priorities into shipped product faster than traditional engineering teams.

While others debate timelines and chase alignment, smart companies deploy engineers who unblock work, accelerate delivery, and reduce iteration cycles.

At Athyna, we provide senior Forward Deployed Engineers who integrate directly into your team:

• Modern tech stack expertise — engineers experienced in Python, TypeScript, React, Node.js, AWS, Kubernetes, Terraform, and more.

• AI-powered matching finds engineers aligned to your stack and use case.

• 4 days from brief to interviews—no hiring drag, no bottlenecks.

• Save up to $80K per hire vs. U.S. talent without compromising seniority or impact.

👉 Meet Your Engineer

(sponsored)

GAMES

POLL RESULTS

What do you think is the best bet for powering AI?

• Geothermal (18%)

• Nuclear/fusion (33%)

• Natural gas (12%)

• Space data centers (15%)

• Something else (write in!) (11%)

• Slow down AI (11%)