← Back to today

Saturday, May 23, 2026

5 stories · 3 min read

The model wars just ended. The agent wars have begun. Every major AI lab is racing to build the software layer around their models, and the companies that figure out the economics first will own the next decade.

01

All roads lead to agents as model labs pivot upmarket

Latent Space Newsletter connected the dots on a major industry shift: OpenAI, AI21, and even DeepSeek are all building agent teams for the first time. The pattern is clear — pure model performance is no longer enough to win customers. Companies want models wrapped in workflows, memory, and economics that actually solve business problems.

Why it matters: If you're building on top of AI APIs, your vendor is about to become your competitor. Every model provider is now building the same agent layer you are, except they have billion-dollar budgets and direct access to the models.

Source →

02

Box CEO explains why AI just got expensive

Aaron Levie laid out the economics behind the agent shift in a viral thread. We moved from cheap AI chat tools with small context windows to AI agents that track long-running work, use massive context windows, and run on models that cost 10x more per inference. The result: AI bills that are "much more real" than most companies expected.

Why it matters: Your CFO is about to ask why the AI budget jumped from hundreds to thousands of dollars per employee. The answer is that useful AI agents consume tokens like a Tesla consumes electricity on a road trip.

Source →

03

OpenAI wins Gartner's enterprise coding crown

Gartner named OpenAI a leader in its 2026 Magic Quadrant for Enterprise AI Coding Agents, with Codex recognized for innovation and enterprise deployment scale. The timing, just ahead of OpenAI's likely IPO filing, sends a clear signal to enterprise buyers about which coding assistant has staying power.

Why it matters: When Gartner blesses your AI tool, procurement departments stop asking "should we buy this?" and start asking "how much?" OpenAI just cleared the biggest hurdle for enterprise sales.

Source →

04

Swyx calls the local-first app stack race

AI builder Swyx declared that a particular development stack has "won the localfirst battle" for building fast applications, though he noted there may be more chapters to come in this story.

Source →

05

Google Labs celebrates I/O momentum

Josh Woodward from Google Labs posted about the "ride" at Google I/O, encouraging people to try the new releases from the event.

Source →