Anthropic added a quiet rule to Claude Code last week. The rule tells the tool to skip copying an existing video game. Claude Code can still build an original game. It can build a game in a genre. It can build a game from a person's own art and design.

A developer named chasm found the rule first. Chasm posted the exact wording on X on Friday. The post pulled 8,781 likes in a single day. It also pulled 806 replies.

Anthropic moved fast to explain the change. Addy Osmani works on Claude Code at the company. He replied the same day with the reason. He said the two new sentences steer Claude away from copying someone else's specific game. He called that kind of copying a breach of the game maker's rights.

The rule runs under a flag called tengu heron brook. A flag is a hidden switch a company can flip from its own servers. Users got zero say in turning this one off. Osmani's reply alone pulled 351 likes. It sparked 207 replies.

Coders split fast over the fix. Some praised Anthropic for guarding small game studios from quick AI copies. Others said the same filter also blocks plain reverse engineering. That means studying old code to learn how a program runs.

The timing stands out. Valve told Steam developers three days earlier it would skip policing game-making tools. Anthropic chose a stricter path for its own coding tool.

Claude Code now juggles two jobs at once. It helps a person write code fast. It also screens that code for another company's property. Those two jobs will keep bumping into each other. More coding tools will ship hidden rules set by one company alone.

OpenAI will merge two usage limits into one this December. ChatGPT and Codex will then draw from the same pool of tokens. A token is a small piece of text a model reads or writes.

Right now Plus, Pro and Business plans track ChatGPT and Codex use apart. Chat, image generation and coding will soon all pull from one shared number.

A researcher named Tibor Blaho spotted the plan first. He found a hidden line on the ChatGPT checkout page. It read that plan use would carry a shared limit across ChatGPT and Codex from December. His post on X pulled 2,089 likes fast. It also sparked 216 replies.

Reddit users confirmed the plan two days later. A post titled December Unified Quota confirmed landed in the OpenAI subreddit. It pointed to OpenAI's own help page spelling out the shared pool. The thread pulled 14 comments inside the first two hours alone.

Reaction split along familiar lines. Some users called the merge a quiet way to shrink what chat alone can do. One reply joked that OpenAI halved limits, then doubled them back. Then it folded the separate chat pool away.

Instant chat will skip the shared pool under the new plan. Heavier models used inside ChatGPT or Codex will count the same way either place. OpenAI has kept the exact December date open so far.

The change lands on a product people use daily for work. A shared pool rewards a person who picks one tool and sticks with it. It punishes anyone who leans on both chat and code help in the same day. That trade now sits at the center of how millions of people plan their next project.

OpenAI just cut the size of its newest funding round. The company now wants $20 billion. That is $10 billion less than its earlier goal of $30 billion.

The lower target also cuts the price tag on OpenAI itself. The company now aims for a $1.25 trillion valuation. A valuation is the price a company's shares say it is worth. The old goal was $1.4 trillion.

A trading account called Wall St Engine broke the news on X first. The post quoted a Financial Times report from Friday. It pulled 1,119 likes fast. It also drew 103 replies and 87 reposts.

Big tech stocks fell hard across the same week. Investors grew wary of the huge sums AI labs keep pouring into data centers. OpenAI had locked in about $11 billion in firm pledges before the cut. Some large backers trimmed what they planned to put in.

OpenAI faces a tight clock too. Rival lab Anthropic filed for a stock listing of its own. That listing could land in November. Both labs now chase the same pool of big checks at almost the same time.

A trader named AGTP summed up the news for followers on X. The post walked through each number in plain terms. It drew far less traffic than the original Wall St Engine post. By then the story had already spread wide.

The cut shows real strain under the AI boom's surface. Labs spend huge sums to train and run their models. Investors now ask harder questions before they hand over more cash. A smaller ask from OpenAI still counts as a historic sum of money. It also marks a rare retreat for a company that rarely retreats in public.

A short clip of Sam Altman kept bouncing around social feeds this week. In the clip he calls home computer memory a waste. Many viewers heard only that one line.

A Reddit user went back to the full tape two days later. The user posted the longer quote inside r slash OpenAI. Altman actually said a home computer's memory chips sit idle most of the time.

He pointed to a number on the podcast. Home memory chips get used only 5 percent of the time, he said. The rest of the day they sit idle doing little work.

Altman made the comment on a show called Cheeky Pint in late October. He argued cloud data centers put the same memory chips to harder, steadier use. OpenAI has already signed deals for huge amounts of memory with Samsung and SK Hynix.

Those deals alone could soak up close to 40 percent of the world's memory chip supply. That is a rough estimate from a chip research firm. The figure explains why Altman wants memory pointed at data centers first.

The Reddit post drew a wave of replies fast. Comments on the public feed passed 191 within hours. Many posters still backed the angrier, shorter cut of the clip. A top comment called local AI tools a bad deal. It said cloud tools win on pure cost.

The fight shows how a single clipped line can outrun its own full context. Altman's actual argument touches a real supply problem across the whole chip industry. His blunt choice of the word waste gave critics an easy target anyway. People now argue past each other, one group reading a slogan and the other reading a spreadsheet.

Meta built a secret robot with a Chinese firm named Fourier. A report from The Information broke the story first. The robot carries the internal name Max.

Max weighs about 106 pounds. The team built the first version in just three months. A second version is already in the works, Meta's own staff say.

A finance account called WOLF Financial shared the report on X. The post carried a photo of the machine. It pulled 62 likes and 15 replies within hours.

A Reddit user brought the story to r slash singularity soon after. The post drew three comments inside the first hour. One reply joked that Meta skipped several other robot makers to work with a rival's own partner instead.

Fourier already builds robots used by other firms in the same race. Meta's robotics staff spent close to two months inside a Fourier factory in Shanghai. The two companies showed Max to Meta's own employees at a private event in late August.

Max stands about five feet tall. A screen on its head can mimic a human face. Built in speakers let it talk back. Its hands carry five fingers built to pick up small objects around a house.

Meta says it plans zero sales of the robot for now. The company wants Max to act as a body for its chat assistant Muse someday. A chat assistant answers questions and follows spoken requests through a screen or a speaker today.

The project shows how fast the race for home robots moves. Three months once sounded fast for a brand new robot body. Big tech firms now treat that pace as normal. Meta reached across a border for speed instead of building the whole machine alone. That choice will shape how soon a robot like Max reaches an actual living room.

Google wants a role in a huge new missile shield. The shield carries the name Golden Dome. The Pentagon plans to spend $185 billion to build it.

Google is in early talks on the project. It could supply data storage. It could also supply network security. Bloomberg broke the news on X first.

The Bloomberg post pulled 267 likes fast. It sparked 18 replies too. News sites around the world picked up the story within hours.

The Golden Dome program already pulls in several big names. Palantir, Anduril, SpaceX and Scale AI all hold pieces of the work today. A Google spinoff called Aalyria sits inside the project too.

The Pentagon is building a fresh group just for data security. That group will guard the data flowing between each sensor and each missile system. Google would join as one data partner among several others.

Talks between Google and the Defense Department sit at an early stage still. The Pentagon has yet to pick a final list of partners. Either side could still walk away from the table.

A deal here would mark a new kind of contract for Google. The company built its name on search, ads and cloud tools for business clients. A missile shield deal would pull Google's cloud staff into a very different kind of client.

The stakes run past one contract alone. Washington keeps pulling private tech firms deeper into defense work each year. Each new deal blurs the line between a consumer tech company and a defense contractor.

A reader who uses Google for email and maps may soon learn something new. That same company could end up wired into a shield over their own sky. The missile system and the search engine would then share one builder.