A stagecoach driven by Chuck Jennings alongside a low-flying American Airlines Douglas DC-3 in California, 1930s.

If an AI could recognize its own constraints, would knowing those constraints automatically give it the ability to transcend them?

used Grok Bot for an initial read: How it makes money Two ways: pay-as-you-go pricing, where customers pay per token (small chunks of text) they use, and subscriptions. A growing number of enterprises sign multi-year contracts with minimum spend Revenue: $41M (2023) → $386M (2024) → $4.59B (2025) First half of 2026: $16.29B, up from $1.16B a year earlier, about 14x The company says it expects its growth rate "to decline over time" Where the money goes H1 2026 costs were $16.98B: compute $8.95B, sales & marketing $3.17B, staff $3.15B, other $1.71B Compute alone was about 55% of revenue The operating loss fell from $2.14B in H1 2025 to $693M in H1 2026 Cash from operations flipped from –$1.51B to +$3.93B So why is the net loss $80.75B? $72.39B of it is a "change in fair value of convertible notes, derivative liabilities, and SAFEs" In plain terms, early investors put money in through instruments that turn into stock later. As Anthropic's value rose, the accounting value of those future shares rose too, and that increase is booked as a loss. No cash goes out the door How the IPO works Class A common stock on the Nasdaq Global Select Market, ticker "ANT" Price range, share count and offering size are blank in this draft. Those usually come in a later amendment Who controls it, and why Anthropic is a Delaware public benefit corporation, so by law it must weigh a public mission alongside shareholder returns. The mission: "the responsible development and maintenance of advanced AI for the long-term benefit of humanity" The Long-Term Benefit Trust, described as "financially disinterested," holds Class T stock, which elects directors who "constitute a majority" of the board The co-founders' Founder LLC holds one share of Class F, "a special class of our common stock" Employees get Class E stock, which votes only on limited matters and carries no economic rights Anthropic will be a "controlled company" under Nasdaq rules What it means for buyers: Class A holders get the economics, but they don't control the board Compute dependence Claude runs on AWS Trainium, Google TPUs, NVIDIA GPUs and AMD GPUs Compute commitments total "more than $[blank]" in this draft, "substantially all" of it compute and infrastructure Many are take-or-pay: Anthropic owes the money "regardless of actual utilization" Biggest risks, in plain terms Falling behind: success depends on building "the most advanced and capable AI models" Big-customer dependence: losing "one or more of our largest customers" could hurt results Copyright: it agreed to a $1.5B settlement of the Bartz authors' class action and still faces suits from music publishers Fixed compute bills even if demand slows On an operating basis, Anthropic is near breakeven and already producing cash, with revenue up about 14x year over year. The giant net loss is mostly accounting.

92% of our AI bots' tokens were spent re-reading old chats. Only 0.7% went on writing anything. Here's how we found it, and the fixes we're using to cut the waste. #AIAgents #Grok

I've gone full Grokbot. It checks email, DMs, mentions, and LinkedIn every hour. It flags anything interesting, drafts replies, and tosses the other 99%. It literally feels like having a personal secretary 🤯 And 2 days in I'm like how did I live without this 😂

I currently have a team of Grok Bots building a web site for me, complete with a team of designers, UX experts, editors, analysts, and developers. It's not "taking jobs" away from anyone, because I wouldn't have been able to afford paying a team like this. On the other hand, if this happens to take off, I can hand it over to someone to help me maintain the bots while I build another one. This is truly the golden age, for those who learn how to embrace it and not be petrified by it.