Feature Request: Persistent Project State — ChatGPT Needs to Remember the Work, Not Just the User

Introduction

I’m not a developer or AI professional. I’m an ordinary user who has started using AI extensively because I can see its potential to become a genuine personal assistant rather than simply a tool for answering questions.

I’ve used ChatGPT for travel planning, research, writing, purchasing decisions, retirement planning and everyday problem solving. I think it is extremely capable, and I can see enormous potential in it.

This proposal comes from a real experience I’ve just had with ChatGPT.

It highlighted what I believe is one of the biggest gaps between an excellent AI chatbot and a genuinely dependable long-term personal assistant.

The problem

ChatGPT needs to reliably remember and maintain ongoing projects.

I don’t simply mean remembering facts about me.

I mean remembering the current state of something we have been working on together — including the decisions we’ve made, changes we’ve made, things we’ve rejected, outstanding decisions and the latest version of the plan.

A real example

I was using ChatGPT to plan a family holiday in Spain.

We built a detailed Summer 2026 Holiday Master over a number of conversations.

The holiday covered two weeks and two different resorts:

  • Week 1 — Calella
  • Week 2 — Salou

Over several days, we progressively built the plan.

We discussed trains, markets, meals, restaurants, transport, activities, PortAventura, costs and changes to individual days.

Rather than repeatedly asking ChatGPT to reproduce the entire plan, I established a simple instruction:

“U” = update the master.

Whenever something changed, I would say “U” and ChatGPT would update the existing master while retaining everything else.

This worked extremely well.

By Sunday, ChatGPT was able to recall the complete two-week master, and we could review it and make further changes.

At that point I wasn’t simply asking questions anymore.

I was working with an assistant that knew what we were doing.

That is exactly how I want to use AI.

Then something went wrong

Later, I asked ChatGPT to show me the master.

Instead of retrieving the existing master, it produced a reconstructed version.

It got fundamental information wrong — including putting the first week in Salou when the first week was actually in Calella.

I challenged it.

It then acknowledged that it didn’t have the complete master available.

The most frustrating part was that it still had fragments of the project and some of the decisions, so it could produce something that looked plausible.

That is actually worse than simply saying:

“I can’t retrieve the master.”

The AI was effectively filling gaps with guesses.

I had deliberately created a system where “U” meant “update the master”, and we had successfully used that system repeatedly.

Yet the actual working document had not been reliably preserved.

Why this matters

This isn’t just a holiday problem.

The same problem exists for:

  • retirement planning
  • buying a house
  • running a business project
  • researching a complicated purchase
  • planning a wedding
  • writing a book
  • managing a renovation
  • planning a long-term trip
  • researching a legal or financial issue
  • learning a subject over many months
  • managing an ongoing work project

Users don’t necessarily want AI to simply remember facts about them.

They want AI to remember what they are working on.

That is a fundamentally different requirement.

What I think ChatGPT needs

I would like to see a proper Persistent Project State system.

When I create a project, I should be able to tell ChatGPT:

“This is the master for this project. Maintain it until I tell you the project is finished.”

ChatGPT should then maintain a persistent, authoritative version of that project’s current state.

For example:

Summer 2026 Holiday — MASTER

Confirmed

  • Week 1: Calella
  • Week 2: Salou
  • Dates
  • Accommodation
  • Transport
  • Activities
  • Meals
  • Bookings
  • Decisions already made

Rejected

  • Options we considered but decided against

Outstanding

  • Things we still need to decide

Changes

  • A record of what has been changed and when

The user should be able to update this from any conversation belonging to that project.

The “U” concept

A simple command such as:

U

could mean:

Retrieve the current project master, apply the latest agreed changes, check for conflicts, and save the updated version.

The important word is current.

ChatGPT shouldn’t try to recreate the project from whatever fragments happen to be most accessible in the current conversation.

It should retrieve the authoritative project state.

Conflict protection

This is extremely important.

Suppose the master says:

Wednesday — Santa Susanna market

and six conversations later the user says:

“Actually, change Wednesday to PortAventura.”

ChatGPT should update Wednesday.

It should not accidentally remove unrelated information about Thursday, Friday or the following week.

If two pieces of information conflict and ChatGPT isn’t sure which is correct, it should ask:

“I have two conflicting versions. Which one should be authoritative?”

It should never silently choose one.

Never invent missing project information

This is perhaps the most important lesson from my experience.

If ChatGPT cannot retrieve part of the master, it should say:

“I can’t retrieve that part of the project state, so I won’t guess.”

A plausible reconstruction is not good enough.

For an ongoing project, accuracy and trust are more important than producing a complete-looking answer.

Version history

I would also like the ability to see:

Master v1
Master v2
Master v3
Current Master

That would make it possible to recover from an accidental change.

It would also give the user confidence that their work isn’t disappearing into an invisible memory system.

Make project memory visible

Users should be able to see what ChatGPT considers the authoritative information for a project.

Something like:

Project Memory

Current master
Decisions
Preferences
Constraints
Important facts
Outstanding questions
Previous versions

Users should be able to:

  • pin information;
  • correct information;
  • delete information;
  • mark information as authoritative;
  • restore an earlier version.

This is especially important because users currently don’t always know what ChatGPT actually remembers and what it doesn’t.

My experience with Microsoft 365 Copilot

I use Microsoft 365 Copilot at work with a premium licence.

It is excellent at writing replies, much like ChatGPT, but what has particularly impressed me is how well it can work within my existing working environment.

For example, its ability to work with Excel is genuinely transformative. It can create formulas that could otherwise take me hours of trial and error.

More importantly, it often feels as though it understands the work and projects I have already been doing.

That comparison has made me realise that the next major step for ChatGPT isn’t necessarily just making the underlying AI more intelligent.

It needs better continuity.

Why I think this could be a major differentiator

ChatGPT already has extraordinary capabilities.

It can research, write, analyse, plan, calculate, explain and solve problems at a level that genuinely saves people time.

But there is a difference between:

“An incredibly intelligent AI that I can talk to”

and

“An incredibly intelligent assistant that knows what we’re working on.”

The second is much more valuable.

I believe reliable persistent project state could be one of the features that turns AI from something people use occasionally into something they genuinely rely on every day.

My proposed principle

I would describe the feature in one sentence:

ChatGPT shouldn’t just remember who I am. It should remember what we’re working on.

And for important projects:

The latest approved project state should be treated as authoritative unless I explicitly change it.

One final point

I don’t think this should necessarily be a Pro-only feature.

It is fundamentally a trust and usability feature, not simply a request for more computing power.

Obviously there could be sensible limits around storage, context and usage, but the basic ability to create an authoritative project state should be available to ordinary users.

I’ve already experienced how powerful ChatGPT becomes when it remembers an evolving project correctly.

The frustrating part is that I also experienced how quickly that value disappears when the continuity breaks.

The technology is already incredibly close. I genuinely believe that reliable persistent project state would be one of the changes that could move ChatGPT from “an amazing AI I use” to “the assistant I genuinely rely on.”

I’m raising this because I genuinely believe in the potential of ChatGPT. I’m not looking for perfection — just a relatively small change in how ongoing work is remembered and maintained could make a huge difference to how people use AI.