Zeitgeist · July 13, 2026
Model Companies Want More of the App Layer
This week, the team kept returning to one concern: the companies that supply the models, chips, and infrastructure may also want the app businesses built on top of them. The crew saw that conflict in talent moves, data access, usage policies, and NVIDIA's support for open models.

Talent moves in both directions
Link: https://x.com/t_blom/status/2076580921398931788
- ▸YC managing partner Tom Blomfield is taking a leave of absence to join Anthropic's compute team.
- ▸The team saw the move as a bet that there is more leverage in models and compute than in the apps built on top of them.
- ▸Anthropic's head of design has reportedly left for Cursor. The crew thought that was a strong hire for a company trying to make AI coding tools easier to use.
Apple and OpenAI's security dispute
Link: https://x.com/i/status/2075795797635956963
- ▸Apple alleges that a former hardware engineer who joined OpenAI kept an Apple laptop and exploited a bug to access its network storage.
- ▸Apple says it discovered a message about the access on its internal system before the conversation moved to Signal.
- ▸The team took the case as evidence that the competition between Apple and OpenAI now extends beyond products and into employees, intellectual property, and security.
When customer data helps a future competitor
Link: https://x.com/satyanadella/status/2076323181154230284
- ▸Satya Nadella's post raised Arrow's information paradox: buyers need to see information to judge its value, but showing it can give away what is being sold.
- ▸The crew worried about that problem when the buyer is also a model provider. Customer context can improve the model, and the provider can use what it learns to enter the customer's market.
- ▸The team distinguished this from ordinary cloud risk. A cloud provider sees usage, while a model provider may learn patterns from the work itself.
Codex drops its usage window
- ▸OpenAI temporarily removed Codex's five-hour usage window while Anthropic users were dealing with outages and tighter limits.
- ▸The team read the timing as a competitive move aimed at frustrated Anthropic users.
- ▸Matthew noted that expiring credits and reset windows encourage people to use more compute before they lose access to it.
The team wants specific AI guardrails
Link: https://www.wemustactnow.ai/
- ▸A statement signed by prominent economists and AI researchers calls for governments and institutions to prepare for AI's economic effects with better incentives and guardrails.
- ▸The crew thought the signatories gave the effort credibility, but found the proposal too broad to evaluate.
- ▸The team wanted specific policies, responsible parties, and measurable outcomes instead of a general call to prepare.
Why NVIDIA supports open models
- ▸NVIDIA benefits when cheaper and more accessible models increase demand for inference and, in turn, its chips.
- ▸The team connected this to Jevons paradox: lowering the cost of a resource can increase total consumption.
- ▸The crew also saw open models as less risky for app builders because they are easier to run independently or replace if a provider becomes a competitor.