Epic Games has told the team on its Disney shooter to stop writing code by hand, and the game no longer has a release date. Character.AI, valued at $1 billion in 2023, has filed for bankruptcy, and Disney is offering $20 million for some of its assets. Grok, Cursor and X Premium now come as one subscription from $8 a month. Here is what each one means for you.

Bloomberg reported on 12 October that a director on Epic Games' unannounced Disney shooter, which features Marvel and Star Wars characters, told the team the project is now "AI first." An internal message seen by Bloomberg said "No more hand-written code" and that the game's art should be "generated with full AI workflows," with developers directing and correcting tools such as Anthropic's Claude. The game had already slipped from November to March before the order, and it now has no release date. Tim Sweeney posted that Epic has "a huge pivot underway" to "retool for maximum AI acceleration."

Why it matters: This shooter comes out of Disney's 2024 deal with Epic, which included a $1.5 billion investment. Sweeney says the goal is for the same people to build more and better games, but Bloomberg says many on the team reacted with confusion and shock. Game developers are split on this: in GDC's annual survey, 52% now say generative AI is bad for the industry, up from 30% in 2025. The project was also running late before the order, so it's hard to say how much of the new delay comes from the AI switch.

What to do:

Source: Engadget

Character.AI filed for Chapter 11 bankruptcy in Delaware on 12 October and agreed to sell some of its assets to Disney for $20 million, subject to court approval. Disney's offer, for a non-exclusive, permanent licence to some of Character.AI's intellectual property, is a starting bid, so other buyers can still come in higher. The company estimates both its assets and its debts at between $50 million and $100 million, and expects to finish selling everything by late November or early December. The app keeps running in the meantime.

Why it matters: Investors valued Character.AI at $1 billion in 2023. Its founders went back to Google in a 2024 licensing deal, and in 2025 Disney sent it a cease-and-desist letter over its use of Disney characters. Now Disney is the first bidder for what's left. If no buyer wants to keep the platform running, the company says it will be wound down.

What to do:

Source: NBC News

On 12 October X launched The X Subscription, which bundles Grok, the Cursor coding tool and X Premium into one plan. It runs from a free tier through Plus at $8, Premium at $30, Super at $100 and Ultra at $200 a month, and replaces buying SuperGrok, Cursor Pro and X Premium separately. Paid tiers share one usage pool across the Grok apps, Grok on X and Cursor.

Why it matters: On paper, $8 a month buys a lot of AI. The catch is on Cursor's own help page: Claude and GPT models inside Cursor aren't covered by the plan and are billed at full API prices from the first request. The bundle is cheap for people happy to code with Grok, and it nudges everyone else toward Grok. Switching is optional, and existing subscribers keep their current plans unless they move.

What to do:

Source: Cursor

In research published on 12 October, Anthropic split a 26-billion-parameter version of Google's Gemma 4 across 15 off-brand Android TV boxes, each costing about $55 and holding 4GB of memory. No single box could hold the model, but together they ran it, slowly. The team was testing how close small AI models are to copying themselves from device to device, and estimates a small model could manage it by the end of 2026, which it calls "only a rough estimate."

Why it's interesting: Critics say the test was set up to succeed. Anthropic made the break-in step easy by leaving the target machines without a password, and Rocky Linux founder Gregory Kurtzer replied that "Fear isn't evidence."

Key takeaway: The cheap smart gadgets that never get updated are the devices this research worries about.

Source: Fortune

Apple launched a new Apple TV 4K on 13 October for $299, its first refresh since 2022, shipping on 13 November. Three far-field microphones let you talk to the updated Siri without picking up the remote, and you can ask things like which Apple TV shows won at the Emmys this year. The remote also gets a small speaker, so you can find it when it slips down the sofa.

Why it's interesting: Bloomberg had reported that Apple planned this box for 2024. Delays to the AI version of Siri held it back for two years.

Key takeaway: Apple's AI Siri is now shipping inside the home devices that waited for it, starting with the TV.

Source: TechCrunch

On 11 October Andrej Karpathy, a founding member of OpenAI, posted a rough funnel of how the world uses AI. By his estimate, about 75% of people have little or no contact with chatbots, around 20% use free tools now and then as a better search engine, and about 20 million people, or 0.2%, use frontier models for serious work in coding and maths. At the bottom he puts roughly 5,000 people with access to the best systems inside the AI labs.

Why it's interesting: He calls the figures napkin math. His argument is that the gap is getting wider, because models improve fastest where answers can be checked automatically, like code.

Key takeaway: If you use AI for real work every week, you're already in a very small group.

Source: Pasquale Pillitteri

Context: Epic made AI the default for a whole game team with one message. Before you set any AI target for your own team, it helps to know which tasks are ready for it.

Prompt: Here are the tasks my team does each week: [list each task, who does it, hours per week, and what a good result looks like]. For each task, say whether AI could produce a usable first draft today, whether it would need a person checking every result, or whether it should stay fully human. Give one sentence of reasoning for each. Then suggest a two-week trial for the three tasks that take the most hours.

Do not: rate a task you can't judge from my description. Mark it as unclear.

If/Then: If a task involves money, contracts or anything sent to customers, keep a person signing it off whatever rating it gets.

Example: A 12-person design agency runs its weekly task list through the prompt. It trials AI first drafts for client status reports and social captions for two weeks, and keeps final artwork with its designers.

CNBC reported that Anthropic will share its third-quarter financials with a group of institutional investors on 14 October, as it prepares for a possible IPO. Anthropic has filed confidentially with the SEC, and no listing date has been confirmed. The company declined to comment.

Why it's important: Investors are watching closely how AI labs count revenue. OpenAI recently told its investors its annual revenue rate was about $50 billion at the end of September, well below the $68 billion that had been widely reported, because the earlier figure included gross revenue from partnerships like its deal with Nvidia. Anthropic counts sales made through its cloud partners, so the two companies' numbers may not compare like for like.

Business takeaway: When you see an AI company's revenue figure, check what it counts before comparing it with a rival's.

Source: CNBC

Pluto Security, which tracks the AI tools and agents employees use at work, announced $42 million in funding on 13 October across a seed and a Series A round. It says about 60% of the risks it detects come from tools vibe-coded by non-developers, with finance teams the biggest source, followed by marketing and legal. Developers come fourth. Pluto has passed $2 million in annual recurring revenue with more than 30 enterprise customers.

Why it's important: These figures come from Pluto's own customers, and Pluto sells the fix. Its CEO also says a new AI tool can show up inside a company within two hours of release, which is faster than most IT teams review anything.

Business takeaway: If people outside your tech team are building their own AI tools, find out what data those tools can reach.

Source: TechCrunch

REA, short for Reverse Engineer Anything, is a free open-source tool that lets AI coding agents examine software without its source code, from desktop and web apps to Android apps, using established tools like Ghidra and IDA. Its GitHub page now shows more than 222,000 stars, up from about 29,900 on 9 October according to one report.

Why it's important: A reviewer at MindStudio used it to rebuild parts of Apple's Preview app and found the copy reproduced the main features reasonably well, though an agent without REA finished faster in one comparison. Copying another app's features keeps getting cheaper.

Business takeaway: If your product's main advantage is a set of features, assume a competitor with an AI agent can copy them, and put more into your data and your customer relationships.

Source: GitHub

That's it for today's Daily Pulse. Forward this to one person who wants to stay ahead of AI. See you in the next one. - Nicolas