# The Frontier — Edition No. 36

> Sunday, November 1, 2026. 7 items from named primary sources, with independent analysis.

## 1. FBI says North Korea was behind the $6.3 billion Binance hack

Source: [The Information](https://www.theinformation.com/briefings/fbi-says-north-korea-behind-6-3-billion-binance-hack)

The FBI said North Korea was behind the $6.3 billion Binance hack, the largest crypto theft ever reported. Binance has pledged to cover 100% of the losses from the breach, which was first disclosed on October 30 after a compromised engineer laptop was identified as a possible entry point.

**The read:** Attribution changes this from a company security failure into a state attack on the financial system. When the thief is a sanctioned government, every exchange, bank, and vendor that touches the stolen funds faces legal and compliance risk, not just a technical cleanup. For founders in crypto and payments, the bar just moved: custody, employee laptop security, and frozen-fund response are now national security questions, and regulators will ask about them before they ask about growth.


## 2. Meta built a humanoid prototype with a Chinese robotics startup, and is already on version two

Source: [The Information](https://www.theinformation.com/articles/meta-built-humanoid-prototype-chinese-startup-ramp-robotics-push)

Meta built a prototype humanoid with Chinese robotics company Fourier and is already working on a second version, as it steps up efforts to develop robots for people's homes. The roughly 106-pound humanoid, known internally as Max, was built in just three months. The goal was to help Meta develop a more polished future product at its robotics arm that it could eventually release.

**The read:** Meta is treating home robots as a product race, not a research project, and it built a first machine in a single quarter by partnering instead of building everything alone. That speed comes with a trade: a partnership with a Chinese firm puts Meta's robotics supply chain and data in the middle of US-China tech tensions. For robotics founders, the signal is that the big platforms will move fast through partnerships, so a startup's edge has to be a part Meta cannot easily buy or borrow, such as dexterity, safety, or a specific home task done better than anyone else.


## 3. Anthropic drops the five-hour limit for verified Claude Ultra users

Source: [Anthropic](https://support.claude.com/en/articles/17728952-what-is-the-ultra-plan)

Anthropic renamed its top Claude plans on October 30: Max 5x is now Max at $100 a month, and Max 20x is now Ultra at $200 a month, with no price change for existing subscribers. Verified Ultra accounts no longer have a five-hour usage limit, starting with most existing subscribers, while the weekly allowance stays at 10 times the Pro plan, with a separate weekly limit for the top Fable model. Newer accounts keep a temporary limit until Anthropic verifies them automatically from account and payment history.

**The read:** The five-hour window was the wall that broke long coding and research sessions, and removing it matters more than the new name. A heavy user can now burn a whole week of allowance in one long run instead of stopping midday, which makes Ultra a real tool for agent work that runs for hours. For builders, the lesson is that labs are competing on how long an agent can keep working without interruption, not just on how smart it is, so plan long jobs around weekly budgets, not session clocks.


## 4. OpenAI will merge ChatGPT and Codex into one shared token allowance in December

Source: [OpenAI](https://help.openai.com/en/articles/20001625-understanding-chatgpt-and-codex-usage-limits)

Starting in December, Plus, Pro, and Business plans will share one usage allowance across ChatGPT and Codex, as OpenAI folds the agent features of Work into the main ChatGPT experience. Usage will be measured in tokens, the units of text a model reads and writes, and more advanced models and harder tasks will use more of the allowance. Instant chat, the fast-response mode, will not draw from the allowance, and combined remaining usage and reset times will show in Settings under Usage. OpenAI gave no exact date and no allowance size per plan.

**The read:** Chat and coding used to have separate budgets, so a heavy coding afternoon could not eat your chat allowance, and the reverse. From December, one pool pays for both, which means teams have to manage a single budget across questions, documents, and agent coding runs. The practical move is to track token burn by task now, because a long agent job will soon compete directly with everyday chat, and the users who run out will have to wait, upgrade, or buy credits.


## 5. a16z warns that falling token prices do not mean falling AI spending

Source: [a16z](https://www.a16z.news/p/charts-of-the-week-job-fud-fades)

In its Charts of the Week for October 30, a16z notes that average token prices have fallen 67% since May, while token spending has still risen roughly 122% to 160%, depending on the data source, even as GPU rental rates rise. The firm calls the gap the Rack Spread, after the spread between crude oil and refined products, but warns the token data comes from small, biased samples of users who expose their spending to vendors, while most large companies do not. The same edition notes that job postings for engineers, software engineers, and manufacturing are up about 30%, 26%, and 13% since last October, and that AI-related layoff announcements in August and September were the lowest of the year.

**The read:** Cheaper tokens are not shrinking the AI bill, because cheaper work gets used far more, especially by automated agents that run long, multi-step jobs. But the spending numbers everyone quotes are built on a thin slice of users, so treating them as proof of demand, or of a bubble, is a mistake. For founders and investors, the right question is not what a token costs, but how much work customers will automate once it is cheap, and whether your own usage data, not a vendor index, shows that growth.


## 6. Box Group's David Tisch: competing with mega-platforms is a losing game

Source: [20VC](https://www.youtube.com/@20VC)

On 20VC on October 31, Box Group founder David Tisch argued that high ownership and concentration in venture funds is overrated, that no manager actually wants to stay small, and that competing head-on with the largest platforms is a trap. Box Group, an early investor in Cursor, Baseten, and Clay, now invests from a $275 million seed fund and a $275 million opportunity fund. Tisch also called the Lean Startup a fake, said funding consumer brands at seed no longer makes sense, and argued that YC has the best business model in venture, after noting that Rogo, now a $2 billion company, was passed on by 39 of 40 investors.

**The read:** Tisch is saying the seed game has changed: winning is less about owning a huge share of one company and more about staying in the best companies as they grow, without fighting giants on their own turf. That is a warning for founders who pitch themselves as a cheaper version of a platform product, and for small funds that try to outbid the biggest names. The companies that survive are the ones that find a corner the giants ignore, prove it with real customers, and keep enough room for follow-on money when the round gets crowded.


## 7. Bending Spoons shows how steady improvement can build a $20 billion company

Source: [a16z](https://a16z.simplecast.com/)

On the a16z Show on November 1, Bending Spoons CEO Luca Ferrari told a16z partner Anish Acharya that systematic improvement can be more dependable than breakthrough invention, when paired with a model built to compound over decades. Bending Spoons buys technology companies, moves them onto a shared platform, improves their economics, and reinvests the cash into new acquisitions. The episode also covers the tradeoffs between growth and profit, how to use debt without putting the company at risk, whether AI threatens software businesses, and how Bending Spoons already uses AI across coding, customer support, and data analysis.

**The read:** Bending Spoons is the opposite of the startup story that chases a single big idea. It buys mature products, runs them better, and uses the profits to buy the next one, which means its advantage is operations, not invention. For founders, that reframes the exit question: a solid, well-run product with real revenue may be worth more to a buyer like this than a flashy prototype with no profit. And for operators, it shows that boring work, done relentlessly across many products, can compound into a company as large as any single breakthrough.


[View this edition](https://thefrontier.news/editions/2026-11-01)