OpenAI has been on a tear this week, rolling out a barrage of product updates in just a few days: Workspace Agents (an upgrade that turns GPTs into team-shared agents), ChatGPT for Clinicians (a free version for doctors), ChatGPT Images 2.0 (2K resolution, dual mode), an open-source local privacy filter called Privacy Filter, Codex's screen-watching memory feature Chronicle, and an open-source Harmony conversation visualization tool called Euphony.
https://twitter.com/jxnlco/status/2047191671860031958
Multiple signals point to GPT-5.5 launching soon. The name briefly appeared at the top of the Codex model selector. Greg Brockman teased on a podcast this week that "a new model is coming on the writing front." The prevailing theory is that OpenAI's next frontier model, internally codenamed "Spud," is GPT-5.5, with pre-training reportedly completed at the end of March.
OpenAI developer experience team member Jason Liu and research scientist roon responded on X today to criticism about the company's release cadence. Some users accused OpenAI of drip-feeding features before "Spud" arrives. Roon replied that models and features ship as soon as they're ready: "It looks like drip-feeding because they were literally finished one by one." Liu confirmed from an insider perspective: before joining, he assumed the company would hold models back to wait for competitors or to bundle features. "Turns out that's not how it works — everything goes out the moment it's done."