Journalism

Highlights #1

GPT-5.5, the war in Iran, Nike job cuts, and more.

It’s April 23, 2026. Today: GPT-5.5, the war in Iran, Nike job cuts, and more.


At noon, OpenAI dropped GPT-5.5, its newest flagship large language model. The release of the model comes almost two months after OpenAI dropped the flagship model GPT-5.4 in March. Since then: The New Yorker released a significant investigation on Sam Altman, OpenAI’s CEO. Anthropic—a competitor to OpenAI—claims to have created a model that’s too powerful to be released to the public. OpenAI released a powerful new model for creating images and claimed it will usher in a “new era of image generation.” Finally, OpenAI turned Codex—an app for developers that allowed them to create code with AI (vibe coding)—into an AI-powered agent for general audiences that can send emails, check messages, and use apps on users’ computers directly.

GPT-5.5 is a good model. OpenAI reports, in particular, that the model is good at “agentic coding, computer use, knowledge work, and early scientific research.” The model scores a 60 on the Artificial Analysis Intelligence Index, three points in front of Claude Opus 4.7—Anthropic’s current flagship model—and GPT-5.4 (which both score a 57). Software developers have been reporting that the model is more efficient and intelligent.

GPT-5.5 is 2x the cost of GPT-5.4, at $5 / 1M input tokens. In my testing, GPT-5.5 did not cost significantly more than GPT-5.4—it looks like the token efficiency improvements cancel out the model’s general price increase. GPT-5.5 is still a significantly cheaper model than Anthropic’s best offerings and poses a real threat to other major AI labs.



FAQs:

What is Highlights?

This is my newsletter. It’s designed to cover a wide range of topics, but focuses primarily on news in politics, business, and tech (AI).

Will this newsletter be politically biased?

Hopefully not. This newsletter aims to be politically neutral when covering political topics. When possible, I will attempt to use a wide variety of sources (that have different political affiliations).

Was AI used in the creation of this issue?

No. I did not utilize AI while creating this issue.