AI Radar for Companies — Wednesday, October 14, 2026
· CompaniesAutomation
Epic Games orders AI-first programming with no handwritten code; Google opens Gemini 4 Argon at half price until November 30, and Omron pays $675M for OnLogic.
Epic Games, the creator of Fortnite, has told the team behind its Disney game to work AI first and stop writing code by hand, and its CEO admits the gains range from 20% to fivefold depending on the task: AI First now has a real-world case study, and its lesson is to measure before you mandate. Google is opening up Gemini 4 Argon at half price, but only through November 30. And the money is moving at two speeds: Omron is paying $675 million for the computers that bring AI to the factory floor, while Character.AI files for bankruptcy and sells some of its assets to Disney for $20 million. This picks up from Tuesday's radar (October 13).
Epic Games orders "no more hand-written code" and delays its Disney game: AI First gets measured task by task, not mandated
Epic Games, owner of Fortnite and the Unreal engine, has delayed, with no new date, the Marvel and Star Wars shooter it's making with Disney, originally slated for March 2027, and has told the team to adopt an "AI first" production model using tools such as Anthropic's Claude, according to internal messages seen by Bloomberg. One of the game's directors put it bluntly: "No more hand-written code," and the art will be "generated with full AI workflows"; people direct and supervise what the models produce. The project, according to the same outlet, was understaffed and moving slower than expected. CEO Tim Sweeney calls it a "huge pivot" meant to let everyone "build MORE and BETTER stuff," not "do the same stuff with fewer people," and he gave his own numbers: gains range from 20% in the most technical parts of Unreal Engine work to 5x in web development. For your business: the first lesson of AI First comes from Sweeney himself: the gains vary wildly from one task to the next. Before declaring anything, time three tasks from three different areas, say a sales proposal, the answer to a common customer question and the monthly report, with and without AI for two weeks. Go AI first where the gain multiplies, not where it tops out at 20%, and put in writing who reviews the AI's output before it goes out. That's exactly the role Epic is keeping for its developers. As we saw with the agent teams that cost thirty times more, adding AI without measuring just multiplies the bill. Declaring AI FIRST in a memo doesn't move you up the ladder: knowing what AI gets you on each task, and who answers for what it delivers, does. Source
Google launches Gemini 4 Argon at half price through November 30: test it now, but run the numbers at the December rate
Google opened Gemini 4 Argon, its most powerful model, to paying API customers and paid Gemini app subscribers on Tuesday, after two weeks of restricted testing. It costs $2 per million input tokens and $10 per million output tokens through November 30; from December 1 the price doubles to $4 and $20, the same price per token as Claude Opus 5.5 and Meta's Muse Wonder that we reported yesterday. Batch jobs, the ones that don't need an immediate answer, cost half, and on the paid tier Google says it doesn't use your data to improve its products. By its own benchmarks, Argon beats GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 in programming, cybersecurity, finance, and law. For your business: six weeks at half price is your chance to run the swap test we keep recommending: your 20 real tasks on the model you use today and on Argon, measuring cost per finished task, not per token. But build the budget on the December price: a task that only pays off at $2 stops paying off on December 1. And anything that isn't urgent, like summarizing your incident history, should go through batch, which stays half price after the promo ends. If your business case only makes sense at the launch price, you're not INTEGRATED: you're running on a promo. Source
Omron pays $675 million for the computers that sit next to the machine: your factory's AI runs on the factory floor
Japan's Omron, one of the big names in industrial automation, has agreed to buy U.S.-based OnLogic for $675 million in cash. OnLogic makes rugged industrial computers, the kind bolted onto the shop floor right next to the line, and sells more than 90% of them in the Americas and Europe. OnLogic had $154 million in revenue in 2025, so Omron is paying more than four times sales; the deal is expected to close around January 2027, with the brand and management team staying in place. Omron wants to go from being just an equipment maker to a factory digitalization and "physical AI" company, combining these computers with its precision control, sensors, and robots, in a year when robotics and physical AI startups have raised roughly $48 billion. For your business: the camera checking parts or the robot picking boxes needs a computer next to the machine, and sector giants are buying these components to sell you the whole package. Before you sign a computer vision or robotics project, ask three things: where the model runs, on the shop floor or in the cloud; whether the line stops when the internet goes down; and who updates that hardware and how often, the lesson from the hacked fridges. And insist on standard hardware, so you can switch software vendors without scrapping the computers. If your line stops every time the internet goes down, your automation isn't UNATTENDED: it's hanging by a cable. Source
Character.AI files for bankruptcy and sells some assets to Disney for $20 million: agree now on how you get your data back if your AI vendor shuts down
Character.AI, the platform for chatting with AI-created characters, has filed for bankruptcy protection in the U.S., with assets and liabilities between $50 million and $100 million, and will sell some assets to Disney for $20 million, plus a perpetual, non-exclusive license for its technology. It was once valued at $1 billion, and in 2024 Google paid it $2.7 billion to license its models and bring back its founders. It will keep running while it looks for a buyer, with the sale expected to close in late November or early December, but its CEO warns there is no bid for the company yet and that, if none comes, the platform will be wound down. The company cites legal, regulatory and operational headwinds, after several lawsuits over the safety of minors. For your business: if your customer service chatbot or agents depend on a startup, its closure or sale could leave you without service and with your data in the hands of whoever buys it. The GDPR already requires your vendor to return or delete personal data when the service ends, but get it in writing, with deadlines: export of your conversations, documents and agent configuration at any time, in a standard format; deletion if the company is sold or shuts down; and advance notice before any shutdown. And keep your instructions and your test set in-house: with those, moving the service to another vendor takes days. If you can't get your data and instructions out in a day, you're not INTEGRATED: you're renting. Source
What to watch tomorrow
Whether Anthropic or OpenAI respond on price to Gemini 4 Argon's launch offer, and whether a buyer for Character.AI turns up. Also, whether Amazon closes its purchase of Decart, which The Wall Street Journal said could happen within days, and what comes out of TechCrunch Disrupt, running in San Francisco through Thursday, October 15.