Writing · a direct comparison
Mneva vs. ChatGPT's memory.
The direct answer: ChatGPT's memory keeps facts about you, quietly, in the background, so a later reply can use them. Mneva keeps the open loop itself — the promise, the decision, the worry — written in your own words, and brings it back to you at the moment it actually matters, not just when you happen to ask.
Where ChatGPT's memory is genuinely good
Credit where it is due, because the honest-blemish rule cuts both ways. OpenAI built memory to be manageable, not hidden: what ChatGPT has saved about you shows up as a list you can open, read, and delete item by item, or clear entirely, inside your settings. You can also turn memory off altogether. That is real legibility, and it is more than a lot of "AI remembers you" products offer. On top of saved memories, ChatGPT can also draw on your past conversations more broadly when that setting is on, which makes it genuinely useful for picking up context across a long relationship with the tool.
Where the two are different
ChatGPT's memory is built to make replies more relevant — it remembers that you have a dog named Otis so a later answer does not make you repeat yourself. That is memory in service of the next reply. It is not built around the idea of an open loop: a thing you said you would do, a decision you are putting off, a worry with no due date. Nothing in a general-purpose memory feature knows that your friend's birthday is the moment your "I should tell her something honest" promise actually matters, or brings it back to you unprompted on that day.
Mneva calls this position Holding. What it keeps, it keeps as a thread you can open and read in the words you used — not a machine's paraphrase filed as a bullet point — and each thread can be edited when you said it badly, closed when it is finished, or deleted outright. When a thread matters again, it surfaces at the moment that makes sense, and either "I did it" or "let it go" closes it for good; nothing nags twice. Read more on this in the essay on holding a promise.
The consent story differs too. Mneva treats anything sensitive — the kind of thing that goes beyond a preference and into how you actually are — as a separate category that runs on your explicit, revocable consent, disclosed inside the conversation itself, not buried in a settings toggle you have to already know exists. The full account, including where limits are, is on the Privacy page.
Side by side
| Question | ChatGPT's memory | Mneva |
|---|---|---|
| Can you read exactly what it kept, in plain language? | Yes — a manageable list in settings. | Yes — a thread in your own words. |
| Does it resurface something unprompted, at the moment it matters? | Not by design; it informs the next reply when asked. | Yes — that is the specific job of Holding. |
| Does an entry ever fully close, with nothing nagging twice? | Not applicable — it is not built around open loops. | Yes — "done" or "let it go" both close it for good. |
| Is sensitive handling gated on separate, explicit consent? | Governed by OpenAI's general privacy controls. | Yes — asked for in the conversation, withdrawable any time. |
An honest limit
Mneva's own live opener sends your message to OpenAI to generate the reply, the same as ChatGPT does — and OpenAI may keep those API prompts and replies in its own safety logs for up to 30 days by default, longer in the exceptions it publishes. Mneva does not control that boundary and does not claim to erase what it never held. What Mneva does control — the thread you can read, edit, and delete, and the consent gate on anything sensitive — is described in full on the Privacy page.