In partnership with
Hey friends 👋
Big weekend for OpenAI, and not entirely in the way they'd choose. GPT-5.6 Sol went public and immediately broke usage records, forcing the company to lift its usage caps just to keep up. Meanwhile Apple filed a lawsuit accusing OpenAI's hardware chief of turning job interviews into an intelligence operation. Let's dig in.
GPT-5.6 Sol Broke Records So Hard OpenAI Had to Turn Off the Limits
OpenAI took GPT-5.6 Sol, Terra, and Luna out of limited preview and into general availability on Thursday, and developers did not ease into it. They jumped straight to the deep end, praising the model's long agentic runs and its new "ultra" mode, which coordinates up to four subagents on a task instead of running everything through a single model instance.
The demand that followed was not subtle. Usage over the launch weekend came in at roughly double the previous record, according to OpenAI's own account of the traffic. Codex and ChatGPT Work now have a combined 6 million active users, a number Codex engineering lead Thibault Sottiaux shared directly rather than through a polished blog post. That's up from 5 million just over a month earlier, which is a strange kind of growth curve for a product category that supposedly plateaus once the novelty wears off.
OpenAI's response to the traffic was to temporarily kill the five-hour usage cap for Plus, Business, and Pro users and reset everyone's quotas, twice, over the weekend. The company also says it's shipping efficiency improvements that should stretch usage further without loosening the cap permanently. None of this required a press release. It happened because the servers were getting hammered and OpenAI needed users to stop hitting a ceiling that was never designed for this much traffic.
Here's the thing worth sitting with: none of the benchmark charts OpenAI published mattered nearly as much this weekend as the fact that people simply would not stop using the thing. Model launches get judged on Terminal-Bench scores for about 48 hours, and then they get judged on whether anyone's still opening the app on day four. Sol cleared that bar loudly.
Speak naturally. Send without fixing.
Wispr Flow turns your voice into clean, professional text you can send the moment you stop talking. Not rough transcription you have to clean up. Actual polished text — ready for email, Slack, or any app.
Speak the way you think. Go on tangents. Change your mind mid-sentence. Flow strips the filler, fixes the grammar, and gives you text that reads like you spent five minutes writing it.
89% of messages sent with zero edits. Millions of professionals use Flow daily, including teams at OpenAI, Vercel, and Clay. Works on Mac, Windows, and iPhone.
Apple Just Accused OpenAI's Hardware Chief of Running Interviews Like Debriefs
Apple filed suit against OpenAI on Friday, and the complaint reads less like a standard trade secrets dispute and more like a corporate espionage file. The named defendants are OpenAI's Chief Hardware Officer Tang Tan, a 24-year Apple veteran who once led iPhone and Apple Watch product design, and Chang Liu, a former Apple engineer who left for OpenAI earlier this year.
The allegations against Tan are specific and strange. Apple claims he directed job candidates who were still working at Apple to bring actual hardware components, things like batteries and logic boards, to "show and tell" interview sessions at OpenAI. One candidate reportedly told Apple he "didn't even know we could take those from the office." Apple also alleges Tan used internal Apple project codenames to pump candidates for details on unreleased products, and that OpenAI circulated a checklist teaching departing Apple employees how to leave without tripping the company's exit security review.
Liu's piece of the complaint is more straightforward but no less serious. Apple says he failed to return his company laptop after leaving, and that he separately gained access to a former colleague's Apple work computer, then downloaded dozens of confidential files covering unreleased products and technical specifications. Apple also says it flagged its concerns to OpenAI back in February and got no response before filing suit.
OpenAI's public reply has been a single line: the company has "no interest in other companies' trade secrets" and remains focused on its own work. That's a fine thing to say. It's a much harder thing to defend line by line in federal court, especially with a named defendant and specific dates attached to the allegations. The timing is brutal too. This lands just as OpenAI is racing toward its own consumer hardware launch and gearing up for a much-discussed IPO, neither of which benefits from a lawsuit alleging your device roadmap runs on someone else's stolen blueprints.
A few more things worth knowing
Cursor shipped version 3.11 with side chats, spin up a parallel conversation with
/sideor/btwwhile your main agent keeps running, then pull the findings back in later with an @-mention. The release also adds searchable agent transcripts and new cloud agent hooks.Side chats in Cursor still count against your usage. They're full agent sessions, not lightweight Q&A, so heavy use will eat into plan limits just like your main thread does.
OpenAI's traffic surge and Apple's lawsuit landed in the same 72-hour window, which either says something about how fast this industry moves or how little breathing room anyone gets between wins and legal exposure.
That's what stood out to me today. Reply and tell me what caught your eye, I read everything.
Talk tomorrow,
Hatman 🎩