OpenAI feature rollout and model rumors
OpenAI's release notes show ChatGPT gaining memory improvements, meeting recording and new connectors to services like Google Drive and SharePoint, while some reports claim a GPT‑5.4 family has been rolled out and that a frontier model codenamed 'Spud' finished pretraining in late March. The official notes confirm product expansions and connector work, but independent reports about GPT‑5.4 and 'Spud' are labelled cautiously in public writeups — making them plausible but unverified rumors about the model roadmap. (help.openai.com) (chatgptimagegenerator.org) (geeky-gadgets.com)
OpenAI’s real April news is less about a mystery model and more about ChatGPT turning into office software: the company’s release notes now show memory upgrades, meeting recording, and more ways to pull in files from work tools like Google Drive and Microsoft SharePoint. The memory piece is simple: ChatGPT is getting better at carrying details from one conversation into the next, so it can remember preferences and past context instead of starting from a blank page every time. OpenAI’s help center describes record mode the same way a human assistant would work after a meeting: it transcribes audio, writes a summary, and saves that summary in chat history for later use. The connector piece is about where ChatGPT can look for information. OpenAI says Pro users can now use chat search connectors for Dropbox, Box, Google Drive, Microsoft OneDrive for Business, and Microsoft SharePoint, which means answers can be grounded in files your company already stores there. OpenAI has also been widening those work links beyond plain file search. Its business and enterprise release notes say ChatGPT can search connected tools like Slack, SharePoint, Google Drive, GitHub, HubSpot, and Asana, while projects can now pull in sources from places like Slack channels and Google Drive folders. That is why the product story and the model story keep getting mixed together. When a new model arrives now, it does not just change the chat box on the website; it can also change coding tools, research tools, company search, and software that acts on your behalf across other apps. On the model side, one part is no longer rumor. OpenAI published an official post on March 5, 2026 saying GPT‑5.4 was being released in ChatGPT as GPT‑5.4 Thinking, in the application programming interface, and in Codex, with a 1 million token context window and native computer-use features in Codex and the application programming interface. That matters because some writeups framed GPT‑5.4 as a leak or whisper campaign even after OpenAI had already posted it publicly. The safer reading is that GPT‑5.4 itself is confirmed, while the confusing part is how fast OpenAI is rolling out variants, product surfaces, and pricing tiers around it. The murkier claim is “Spud.” OpenAI has not published a product page or release note for a frontier model with that codename, and the public articles pushing the story rely on secondary reporting, unnamed sourcing, or rumor roundups rather than a direct OpenAI announcement. So the clean version of this story is: the feature rollout is real, the connector expansion is real, meeting recording is real, and GPT‑5.4 is officially announced. The idea that a separate model codenamed Spud finished pretraining in late March may be plausible, but as of April 10, 2026, it still sits in the bucket of unverified roadmap chatter rather than confirmed OpenAI product news.