ChatGPT tests personality modes: Cynic, Robot, and study tutor

ChatGPT might soon sound a whole lot less generic.

OpenAI is rolling out personality toggles for ChatGPT, spotted by researcher Tibor Blaho under the Customize ChatGPT menu. Instead of its usual diplomatic default, users can soon pick from four distinct personas.

There is Robot for direct, no-nonsense replies; Cynic for sarcastic and critical feedback; Listener for thoughtful and supportive chats; and Sage for an enthusiastic, knowledge-heavy tone. It is rolling out gradually, so not everyone has it yet.

Alongside the personalities, OpenAI is also introducing a Study Together feature. Rather than spitting out instant answers, this mode breaks down complex concepts step by step, introduces topic-specific quizzes, and helps users prep for exams in an interactive format.

Sora 2 and GPT-5 leaks surface ahead of DevDay

While the frontend gets personal, OpenAI's backend is showing early traces of its next major AI models.

Recent server updates point to an internal model named Sora 2. The original Sora launched back in December 2024, made waves, and then largely faded as competitors advanced. Most notably, Google recently made its Veo 3 video generator free for students via Gemini AI Pro, ramping up pressure on OpenAI to respond.

Meanwhile, references to GPT-5-reasoning-alpha have also surfaced in testing. OpenAI's annual DevDay is set for October 6 in San Francisco, which could serve as the launchpad for a GPT-5 preview.

Compute shortages have heavily bottlenecked development, though. Hardware constraints previously forced OpenAI to tightly restrict access to GPT-4.5 earlier this year, leaving GPT-4o as the everyday daily driver for most people. While waiting on bigger infrastructure, the company recently rolled out specialized research models like o3-deep-research and o4-mini-deep-research.

Why it matters

For regular users and students, these quality-of-life updates make ChatGPT dramatically more usable today. You no longer need elaborate custom system prompts just to stop ChatGPT from sounding like an eager customer support rep, and guided tutoring is far more valuable for actual learning than having AI do your homework for you.

For the broader AI landscape, this shows OpenAI playing defense on the product layer while compute bottlenecks hold back its next giant leap. Google has been aggressively bundling video and study tools across its ecosystem, meaning OpenAI cannot afford to let ChatGPT feel stale.

We think shifting focus toward practical personalities and guided workflows is a smart way to squeeze value out of existing models without burning massive server capacity. The risk, however, is that users are eager for real architectural progress. If DevDay comes and goes without a clear, scalable roadmap for GPT-5 and Sora 2, OpenAI risks losing its frontier crown to rivals moving faster on multimodal infrastructure.