Hello OpenAI Team and Community,
I wanted to take a moment to share some positive feedback and a few thoughts about where I would love to see ChatGPT go in the future.
I have been using ChatGPT since its early versions, and over the years it has gradually changed from something I tried out of curiosity into a tool that I use constantly in everyday life. I am also a paying subscriber, and for me the value of the subscription is very easy to justify: ChatGPT saves me an enormous amount of time.
A simple example is troubleshooting. If something stops working on my computer, I could certainly search Google, read Reddit threads, browse support forums, compare solutions, copy commands, and eventually solve the problem myself. But that might take an hour. With ChatGPT, I can describe what is happening, show the results of each step, and work through the problem interactively.
The same applies to researching products, understanding technical information, comparing options, working with documents, or learning difficult material. The information often exists somewhere on the Internet. What ChatGPT gives me is the ability to find, organize, explain, and apply that information without spending hours doing the intermediate work myself.
One recent improvement deserves special thanks: continuity and memory across conversations.
For a long time, one of my biggest frustrations was having to find an old conversation or explain the same context again whenever I started a new chat. Recently, I noticed that ChatGPT could naturally pick up relevant details from previous conversations. It may sound like a small feature on paper, but as a long-time user, it completely changes the experience. It makes ChatGPT feel much more like one continuous assistant rather than hundreds of disconnected conversations.
Please continue developing this. For me, this kind of improvement is exactly what makes a paid subscription worthwhile.
I would also like to share what I personally imagine as the next major step for ChatGPT on a computer, particularly on a Mac.
I would love to see ChatGPT evolve beyond being an application that I open into something closer to a permission-based intelligence layer for the operating system.
The important word is permission-based. I would not want an AI to automatically have access to my entire computer. Instead, I would like to decide exactly which folders, files, applications, photos, email, browser data, etc. it is allowed to access, and what it is allowed to do with them.
For example, after many years of using computers, I have accumulated an enormous amount of digital clutter: documents, PDFs, photos, videos, duplicate files, old downloads, and poorly organized folders. I have wanted to organize it for years, but I simply never have enough time.
I would love to be able to tell ChatGPT:
“I am giving you access to this folder. Analyze it, find duplicates, organize my documents, sort my photographs and videos, and propose a better structure.”
For photographs, I might say:
“Put photos from these years into separate folders. Find pictures where I am skiing, pictures from the lake, pictures with my dog, and organize them into appropriate categories.”
I would then watch the process and correct the AI when necessary. Over time, it could learn how I prefer my files to be organized.
However, I believe destructive actions should remain under human control.
Instead of deleting files that ChatGPT believes are unnecessary, it could move them into a dedicated review folder — something like “AI Cleanup — Review Before Deleting.” After the organization is complete, ChatGPT could tell me:
“I organized 20,000 files. I found 1,500 probable duplicates and unnecessary files totaling 40 GB. They have been moved to the review folder. Nothing has been permanently deleted.”
I could then review them and decide what should actually be removed.
A complete Undo function for an organization session would also make this much safer. AI should be able to handle the repetitive work autonomously while irreversible decisions remain under the user’s control.
Another example would be documents.
I would like to be able to say:
“Find the PDF with this name, open it, make these changes, save the corrected version, and prepare an email with the document attached.”
Or:
“Find all of my statements from 2026, organize them by month, and tell me which months are missing.”
The goal is not simply for ChatGPT to answer questions about my computer. I would like to delegate routine computer work to it.
Voice control could make this even more powerful.
I imagine opening my Mac and simply saying:
“Open this news website and tell me what the major stories are today.”
ChatGPT could summarize the page verbally. I could then say:
“Open the third story and tell me more.”
Then:
“What are other reputable sources saying about the same event?”
Or:
“Save this article, summarize it, and put the summary in this folder.”
All of this could happen as one continuous conversation without requiring me to constantly switch between the keyboard, mouse, browser, Finder, and ChatGPT.
In other words, I do not think voice should only be another way to have a conversation with an AI. I think voice could eventually become an interface for controlling the same AI agent that can, with permission, interact with the computer.
I also think a hybrid local/cloud architecture could be very interesting.
A local component could potentially handle tasks that do not require a massive cloud model: indexing files, reading metadata, finding exact duplicates, organizing folders, performing approved file operations, and perhaps some local classification.
The cloud model could then be used when more sophisticated intelligence is required — complex reasoning, understanding difficult documents, web research, or tasks that genuinely benefit from a large model.
Ideally, the system would send only the information necessary for a particular cloud task rather than requiring an entire personal archive to be uploaded.
This could potentially benefit everyone: users could gain greater privacy, responsiveness, and control, while OpenAI could potentially avoid using expensive cloud compute for operations that a modern computer can efficiently perform locally.
I am not an AI engineer, so I am not suggesting that this is necessarily the correct technical architecture. I am simply describing the experience I would like to have as a user: local computing where it makes sense, cloud intelligence where it is needed, and clear permissions connecting the two.
Ultimately, I imagine ChatGPT becoming less like another program on my computer and more like a trusted personal computer agent.
I tell it what outcome I want. It handles the repetitive steps. I can watch what it is doing, interrupt it, correct it, approve sensitive actions, and remain in control.
There will always be specialized professional software and tasks that require human expertise. I am not asking ChatGPT to replace every application or every professional. I am describing ordinary everyday computing: managing files, finding information, organizing years of digital clutter, working with documents, browsing the web, summarizing information, and connecting these tasks together.
That is where I believe an AI assistant could save an extraordinary amount of human time.
Finally, I simply want to say thank you.
I have watched ChatGPT evolve from the early versions into something that has become genuinely useful in my everyday life. Improvements like better continuity between conversations may not always be the most spectacular features to advertise, but for a long-time user they can fundamentally change how the product feels.
I am happy to pay for a product that saves me time and continues to improve. I hope OpenAI keeps pushing in this direction while keeping the product accessible and reasonably priced.
Thank you to the engineers, researchers, designers, and everyone else working on ChatGPT. Your work is noticed and appreciated by the people who use it every day.
Best regards