ChatGPT And Codex Are Merging (This Changes Everything)
TL;DR
Codex and ChatGPT are becoming one product: OpenAI said it will put Codex into ChatGPT in the next few weeks so users stop choosing between separate apps and instead get one unified agent experience everywhere they work.
OpenAI is pushing from chat toward persistent goal-based agents: Riley highlights Codex's /goal feature and OpenAI's cloud-first vision where agents keep working 24/7, using context from email, calendar, Slack, and iMessage to act proactively instead of waiting for prompts.
Six new role-specific plugins package context, domain knowledge, and workflows: OpenAI showed plugins for sales, data analytics, creative production, product design, public equity investing, and investment banking, with connections to systems like Salesforce, Databricks, and Snowflake.
Annotations turn AI output into a live collaborative surface: Instead of rewriting prompts, users can click directly on spreadsheets, decks, documents, and app previews to ask Codex to explain, edit, or transform specific parts, similar to comments in Figma or Google Docs.
'Sites' is the sleeper announcement with the biggest product implications: OpenAI's early preview lets teams turn Codex output into secure sharable dashboards, prototypes, tools, or apps, which Riley sees as a direct threat to platforms like Lovable and Replit over time.
Riley's core claim is that AI is erasing skill boundaries inside companies: His favorite part of the demo is that anyone can try work outside their role, whether a non-technical person building a production-ready app or a technical person generating marketing assets in minutes.
The Breakdown
OpenAI says Codex is coming into ChatGPT within weeks, turning separate chat and coding products into one agent platform that works across desktop, mobile, browser, and tools like Slack, Excel, and PowerPoint. Riley Brown argues this points to a bigger shift: always-on cloud agents, role-specific plugins, editable annotations, and sharable mini-apps that could put pressure on startups from Harvey to Lovable and Replit.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.

Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
The Art of Tasteful Prompting
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.