Somewhere in San Francisco this week, lawyers at a company that put the word “Open” in its name are drafting paperwork against the company that put their product on a billion phones. OpenAI is reportedly preparing legal action against Apple. That sentence would have been comedy a year ago. Today it’s just Thursday.
It is worth sitting with how strange the turn is. The original pitch of every consumer-facing AI lab was a kind of generosity — a shared frontier, a public good, a tool too powerful to belong to anyone in particular. The pitch made sense when nobody had a business model yet. Now everyone does, and the business model is the same one every platform eventually arrives at: own the surface, charge the rent, make sure no other surface can replace you.
So $AAPL, which controls the surface where most people will first meet AI, and OpenAI, which controls the AI most of those people will reach for, were never going to coexist peacefully forever. The partnership was a marriage between two parties who each privately believed they were the more important half. The dispute, in whatever form it takes, is just that disagreement getting expensive.
Meanwhile, on a different floor of the same industry, OpenAI is rolling out a personal finance product that asks you to connect your bank accounts. Read that twice. The same company whose name is allegedly worth fighting over now wants to know your balances, your recurring bills, what you spend on groceries. Not because it is malicious. Because if a model can answer where your money is going, it can probably answer where it should go — and a model that answers that question routes capital, which is a much larger business than answering questions about Python.
This is the move. AI as the front door to every adjacent industry. Not search, not chat — finance, health, scheduling, shopping. The chat box was the demo. The bank account is the product.
What the exits are telling you
While the partnerships fracture at the top, the org charts fracture at the bottom. Musk’s SpaceXAI has been quietly bleeding staff since the merger that combined the two halves. It rarely makes headlines because nobody leaves with a manifesto; they just leave. One by one. The merger was supposed to consolidate engineering talent. What it consolidated was attrition.
There is a pattern here you can find without squinting. The big consumer-facing AI consolidations of the last eighteen months — by acquisition, by partnership, by forced merger — were sold on synergy. What they delivered was friction. The interesting people on either side discover they now report to a person who used to be a peer’s peer, and they find an excuse to leave. The org chart wins. The output suffers. Investors are slow to notice because revenue is a lagging indicator and morale is invisible on a quarterly call.
The boring bet
Here is the part most people aren’t watching. While the giants spend the week negotiating who gets to own the customer relationship, somebody shipped a small project called Sx — an open-source package manager for AI skills, MCPs, and the small commands that increasingly do the actual work. Think of it as the equivalent of apt-get for the thing you actually use AI for. It is a quiet repository. It will not be on the cover of anything.
But package managers are how industries get cemented. Whoever owns the boring distribution layer underneath the exciting product layer ends up with disproportionate leverage, because every developer eventually has to use that boring layer to ship anything. The web has npm. Linux has apt. Mobile has the App Store. AI’s package manager hasn’t been chosen yet, and the slot is open in a way most slots haven’t been in a decade.
The companies elbowing each other this week are fighting over who owns the storefront. The quiet projects are deciding what gets shelved on it.
The clean-energy footnote
A side note worth keeping in peripheral vision. The tweet thread cycling through $TSLA, $RIVN, and $LCID this week as the “powering the future” trio is doing a familiar thing. Bundle the survivor with the strugglers, suggest a category trade, hope the category does the lifting. The category is real. The basket is not. $TSLA at this point is closer to an energy-and-robotics holding company than a car maker; $RIVN is a capital-intensive bet on one specific truck market; $LCID is a luxury experiment kept alive by foreign capital. Treating them as a single trade is treating “the future” as a single trade, which is the kind of thing people do when they are trying to sell you a basket.
What it adds up to
The companies that promised to be open are calling lawyers. The companies that promised to merge are losing the people who made the merger worth doing. The companies that promised to power the future are being marketed in three-ticker bundles. Underneath all of it, a small project just published a package manager nobody is talking about yet.
This is what an industry looks like when it stops being a frontier and starts being a market. The romance recedes; the rent-seeking begins. The frontier doesn’t disappear — it moves underneath, into the boring layers, where the next decade’s leverage is quietly being decided by whoever bothers to ship the unsexy plumbing.
The headlines will keep being about the storefront. The history will be about the package manager.
Leave a Reply