OpenAI has a habit of making headlines with its blockbuster announcements — GPT-4o, the voice mode overhaul, the Sora video generator. But between those marquee launches, the company has been steadily shipping smaller updates to ChatGPT that have significant practical implications for the millions of people who use the tool daily. Several of these features arrived with little fanfare, buried in changelogs or rolled out gradually across user tiers, and many subscribers still don’t know they exist.
The pace of iteration has accelerated in 2025, with OpenAI pushing updates on what sometimes feels like a weekly cadence. For power users, enterprise customers, and developers building on top of the platform, keeping track of what’s new has become a job in itself. Here are three recent additions that deserve far more attention than they’ve received.
Canvas Mode Graduates From Experiment to Everyday Tool
One of the most consequential but underappreciated additions to ChatGPT is the maturation of Canvas, a feature that gives users a dedicated workspace for writing and coding directly alongside the chat interface. As MakeUseOf reported, Canvas has evolved from a limited beta experiment into a genuinely useful collaborative environment that changes how people interact with AI-generated text and code.
Rather than receiving a block of text in the chat window and then asking for revisions through follow-up prompts, Canvas opens a side panel where users can highlight specific sections, request targeted edits, and watch changes happen in place. For writers, this means you can ask ChatGPT to adjust the tone of a single paragraph without regenerating an entire article. For developers, it means you can point to a specific function and ask for optimization or debugging without losing the context of the broader codebase.
Why Canvas Matters More Than It Seems
The significance of Canvas goes beyond convenience. It represents a fundamental shift in the interaction model between human and AI. Traditional chatbot interfaces are linear — you type, the model responds, you type again. Canvas introduces spatial reasoning into the workflow. Users can work on a document the way they would in Google Docs or VS Code, but with an AI collaborator that understands the full context of what’s on screen.
OpenAI has also added inline suggestions and the ability to control the reading level of text within Canvas, features that make it particularly valuable for educators, content strategists, and technical writers. The tool now supports Python and JavaScript code execution within the panel, allowing developers to test snippets without switching to a separate environment. Despite these capabilities, many ChatGPT subscribers — including those on the $20-per-month Plus plan — have never opened Canvas, according to discussions across developer forums and social media platforms like X.
Scheduled Tasks Turn ChatGPT Into a Persistent Assistant
Perhaps the most quietly transformative feature OpenAI has introduced is the ability to schedule recurring tasks within ChatGPT. As MakeUseOf noted, this feature allows users to set up prompts that execute on a schedule — daily briefings, weekly reports, regular data checks — without needing to manually open the app and type the same request repeatedly.
This is a notable departure from the on-demand nature of chatbots. Until now, every interaction with ChatGPT required a human to initiate it. Scheduled tasks flip that dynamic. You can instruct ChatGPT to send you a morning summary of the top news in your industry, a weekly digest of changes to a competitor’s website, or a daily motivational prompt — and it will do so automatically at the time you specify.
The Enterprise Implications Are Substantial
For business users, the scheduling capability opens up workflows that previously required dedicated automation tools like Zapier or custom scripts. A marketing manager could schedule ChatGPT to generate a weekly performance summary based on predefined metrics. A project manager could have it send Monday morning status update templates to the team. The feature is still relatively basic compared to full-fledged automation platforms, but its integration directly into ChatGPT lowers the barrier to entry dramatically.
There are limitations, of course. Scheduled tasks currently work within ChatGPT’s own interface and notification system, meaning they don’t yet plug directly into email, Slack, or other communication tools without additional setup. But the direction is clear: OpenAI is positioning ChatGPT not just as a tool you consult, but as a persistent agent that works on your behalf even when you’re not actively engaged with it. This aligns with CEO Sam Altman’s repeated public statements about the company’s ambition to build AI agents that can take actions autonomously.
Memory and Personalization Get Sharper
The third feature that has flown under the radar is the continued refinement of ChatGPT’s memory system. OpenAI first introduced memory in early 2024, allowing ChatGPT to remember details from previous conversations and apply them to future interactions. But the initial rollout was inconsistent — the model would sometimes forget things it had been told to remember, or recall details inaccurately.
Recent updates have made memory significantly more reliable and transparent. Users can now view, edit, and delete specific memories through a dedicated settings panel. More importantly, ChatGPT has become better at knowing when to apply stored memories and when to treat a conversation as a fresh interaction. As MakeUseOf highlighted, this improved memory function means the AI can maintain context about your preferences, writing style, job role, and ongoing projects across sessions.
Personalization as Competitive Advantage
This matters because personalization is becoming a key differentiator in the AI assistant market. Google’s Gemini, Anthropic’s Claude, and Microsoft’s Copilot are all competing for the same users, and the assistant that best understands an individual’s needs over time will likely win their loyalty. Memory is the mechanism through which that understanding develops.
For enterprise customers, memory has additional implications around data governance and privacy. OpenAI has been careful to give users full control over what ChatGPT remembers, and the company has stated that memories are not used to train its models. Still, the feature raises questions that IT departments and compliance teams will need to address, particularly in regulated industries like finance and healthcare where data retention policies are strict.
The Bigger Picture: Incremental Updates, Compounding Impact
Taken individually, none of these features — Canvas, scheduled tasks, improved memory — would dominate a news cycle. They lack the visual drama of a new model launch or the controversy of an AI-generated deepfake. But collectively, they represent a significant expansion of what ChatGPT can do and how it fits into daily workflows.
OpenAI’s strategy appears to be one of steady accretion. Each update makes ChatGPT slightly more capable, slightly more personalized, and slightly more autonomous. Over time, these increments compound. The ChatGPT of today is a meaningfully different product from the one that launched in November 2022, not because of any single update, but because of dozens of small ones layered on top of each other.
What Users Should Do Now
For subscribers who haven’t explored these features, the recommendation is straightforward: open Canvas the next time you’re working on a document or piece of code, set up at least one scheduled task to test the automation capability, and review your memory settings to ensure ChatGPT is retaining the right information. These features are available to Plus and Team subscribers, with some also accessible on the free tier in limited form.
The AI assistant market is moving quickly, and the companies that win will be those whose products become so embedded in users’ routines that switching costs become prohibitive. With Canvas, scheduled tasks, and improved memory, OpenAI is making a quiet but deliberate play to ensure ChatGPT becomes not just a tool people use occasionally, but one they rely on continuously. The features may have arrived without fanfare, but their long-term impact on how people work with AI could be profound.