Oct 20, 2026

AI Weekly: The Model Is the Swappable Part

Atlassian, Uber and LangChain treat the AI model as replaceable and guard the gateway and harness around it. Chinese open models stay the hard call.

Mike Cannon-Brookes gets two kinds of questions from Atlassian's enterprise customers. One kind he can satisfy. "there is certainly some customer pushback in some places to quote unquote Chinese models. There is not customer pushback to open models." So Atlassian runs open models, and draws one line: "And so we don't use any Chinese models in production."

He does not expect the line to hold everywhere. "But for an enterprise to say, I'm not gonna use a cheaper alternative that'll give me better results, when I know what's in it, I think it's gonna be a very hard, long term choice". Plenty of companies already cross it quietly. The model a startup calls its own, he says, is often "a Chinese model that they've trained or added their own fine tuning specialization on top of, and then they call it their own model." His job, as he describes it: "It's not to pick a model. It's to use the best model for the best job".

Oct 19, 2026

SaaS Says: Revenue per Compute

Software leaders say headcount no longer measures output. A former Netflix CFO wants revenue per compute on the scorecard next to revenue per head.

Jim Cook got a call from a client last week. "I just had one the other day with one of my clients that got surprised by a $500,000 a month bill on compute."

Cook, a former Netflix CFO, has seen this before. Cloud in 2013: engineers got the power to spend and finance had no language for it. "And the answer is no controls. We don't have controls because we don't even have the definition of talking about or this concept that compute costs money." His fix is not new. "I fundamentally believe it's applying what we already know to a different problem. We've already had to tag cloud costs." Then he moves the scorecard. "We all want our leaders thinking about the cost of their business, but we used to think about it in terms of headcount cost." His replacement: "I'm going to start talking to you about revenue per compute, not revenue per head count."

Oct 16, 2026

Weekend Listens: Plan for the Bad Day

Three interviews on the failure case: private equity exits that stall, DNS records nobody owns at Block, and who pays when cargo is stolen in Mexico.

Oct 15, 2026

Proof a buyer can check

Private equity returns about 10% of NAV a year to LPs and its premium is thinner. AI shows up as proof a buyer can check, like a 36-hour close.

Oct 14, 2026

Your next customer is an agent

Personal AI agents now research and buy for people. They see no ads and show up in no analytics. Brands need to be the answer the agent picks.

Oct 14, 2026

PE Beat: The Token Phone Call

AI spend in PE-backed companies has gone from celebrated to audited, and one CFO expects it in quality of earnings reports at exit.

Oct 13, 2026

AI Weekly: The Card on File

Personal AI agents are only as useful as the access they hold. Investors and operators weigh who answers when an agent spends, signs, or errs.