In the wake of OpenAI’s exciting 12-day Shipmas celebration, CEO Sam Altman engaged with users on X, sparking lively conversation by asking, “What would you like OpenAI to build or fix in 2025?” As you might expect, the post quickly caught fire, with users eagerly sharing their thoughts on future AI innovations.
Right from the get-go, a user chimed in with the idea of a drag-and-drop user interface that allows for multiple models to be linked together. This, they believed, could enhance and refine outputs by delegating tasks to models that specialize in specific areas. This concept had already been buzzing in OpenAI’s Developer Forum, where one comment suggested, “You might see better outcomes by breaking the task into several focused parts.” Altman himself found this suggestion quite intriguing, admitting he hadn’t thought of it before.
Another participant didn’t hold back, presenting a wish list that included a more powerful successor to GPT-4o and the seamless integration of GPT/o/Sora into ChatGPT. They also hoped for extended context, frequent updates, and a “middle ground” subscription plan priced around $50 to $70. While somewhat enigmatic, Altman hinted that such users will likely be pleased with OpenAI’s developments throughout the year.
Interestingly, a call was made to relax the restrictions on the next wave of AI models with a “grown up mode”—suggesting a more flexible approach to development. Altman acknowledged this need for a bit more maturity in their AI offerings.
One parent requested family accounts, enabling them to maintain appropriate guardrails while letting their children’s curiosity flourish. The idea was to combine open exploration with sensible parental controls to avoid things getting out of hand.
X user @davidlach asked about proper agents, prompting Altman to cheerfully reply, “Happy 2025.” Rumor has it that OpenAI plans to roll out an AI assistant called Operator in January 2025. Last year, Kevin Weil, OpenAI’s Chief Product Officer, predicted that “2025 is the year agentic systems will truly go mainstream.”
Now, what can we expect from OpenAI in 2025 and beyond? It looks like we can anticipate a follow-up to the GPT-4o with enhanced capabilities. OpenAI is working on AI models that provide longer memory for generating responses and perhaps a “grown up mode” with fewer constraints, though this development stirs up concerns about safety and privacy. Sora might get a major boost, potentially expanding its features and video capabilities.
The potential of AGI seemed to be a focal point, with speculation swirling that OpenAI might be close to reaching this milestone—or even surpassing it—thanks to the o1 reasoning model. Another report hinted that OpenAI could be considering changes to a strict AGI clause, possibly affecting its partnership with Microsoft to secure further investments and access to cloud resources.
Altman has suggested that AGI could be feasible with today’s technology in the next five years, with surprisingly minimal disruption to society. As he mentioned, “Many great updates we have coming were mentioned not at all or very little!” So, it seems like OpenAI has some thrilling surprises up its sleeve for 2025, and the community is buzzing with anticipation.
