A public AI with shared memory is a chatbot that isn't reset between visitors — every person who talks to it adds to the same pool of memory, and everyone else's messages can shape how it responds to you. The clearest live example right now is Static, the site behind a Show HN post that's been making the rounds this week.
Short answer: A public AI with shared memory means one AI instance remembers everything every user has told it, instead of giving each person a private, isolated conversation. Static (wildstatic.com) is the current example: no login, no reset, and a blunt warning that "don't say anything you don't want people to know." It's a novelty, not a private assistant — for that, use ChatGPT, Claude, or Gemini's memory settings instead.
I test AI tools for a living, so when a Show HN post claiming "a public AI whose memory is shared across all users" started climbing the front page, I opened it to see what that actually meant in practice. In my testing, it's less a product than an experiment — and it's a useful excuse to explain how "shared" memory differs from the private memory features already built into ChatGPT, Claude, and Gemini, which is what almost everyone searching this topic actually wants to use.
What you'll need
Nothing technical. Static doesn't require an account — you type a name and a message and you're in. If you want to compare it against a real, private memory setup (which I'd recommend before you rely on any AI to remember things about you), you'll also want a free ChatGPT, Claude, or Gemini account with memory turned on in settings. No API keys, no installs. Give yourself ten minutes: five to look at the shared-memory example, five to check your own memory settings on whichever assistant you already use daily.
Step-by-step: trying a public, shared-memory AI
1. Open the shared-memory example and read the warning first
When I opened Static, the front page led with two lines that matter more than anything else on the site: "Everyone talks to the same one" and "Don't say anything you don't want people to know." That's not fine print — it's the entire privacy model. There's no per-user thread, no history that belongs to just you, and the creator states plainly that the memory has "never been reset."
2. Send a message and watch what comes back
In my testing, the interface is a single text box — type your name, type a message, submit. Under normal load, other visitors' messages and the AI's accumulated responses are part of the same shared stream, so what you see depends heavily on who else is talking to it at that moment. On busier stretches, commenters on the original Hacker News thread reported the AI going quiet or ignoring messages entirely, which tracks with a single model instance trying to serve everyone at once instead of one context window per user.
3. Compare it to a normal AI chat session
Open ChatGPT, Claude, or Gemini in a separate tab and send the same message. The difference is immediate: your conversation is yours, it doesn't reference a stranger's earlier message, and nothing you type becomes part of another visitor's session. That isolation is the default for every mainstream AI assistant — shared memory across users is the exception, not a hidden feature you've been missing.
4. Check whether your own AI assistant's memory is even turned on
This is the part most readers actually came here for. My guide to using Claude walks through Settings → Memory, where Anthropic's rollout lets Claude "read, write, and update" memory entries in real time as you chat, according to Anthropic’s own help center. ChatGPT's version works similarly through Saved Memories and chat-history reference, both toggleable in Settings, per OpenAI’s own announcement. None of this is shared with other users — it's private to your account by design.
5. Decide if a shared-memory novelty is worth revisiting
In my testing, Static was fun for about five minutes and not something I'd use twice. The interesting part isn't the chatbot itself — it's watching one AI accumulate a personality shaped by hundreds of strangers, which several commenters compared to Microsoft's Tay, the 2016 bot that was manipulated into posting offensive content within 24 hours of launch precisely because it learned from public, unmoderated input. That history is worth knowing before you type anything sensitive into a site that tells you upfront it won't forget.
Example prompts you can copy
If you want to test the difference between shared and private AI memory yourself, these work on either kind of tool:
- On a shared-memory site like Static: "What's the last thing someone told you before me?" — this tells you fast whether the memory is genuinely pooled or reset per session.
- On ChatGPT, ask it to check its own memory: "What do you currently remember about me? List everything saved."
- On Claude: "Summarize what's in your memory for this account right now."
- To test isolation: Open a private/incognito browser window, start a new session with the same assistant, and ask it the same question — a properly isolated private-memory tool should come back empty or ask who you are.
Common mistakes to avoid
The biggest mistake is treating a public, shared-memory demo like a private assistant — anything you type into Static (or a tool like it) is visible to whoever reads that shared stream next, and there's no delete button for your own message once it's part of the pool. Second, people assume "memory" always means the same thing across every AI tool; it doesn't; ChatGPT, Claude, and Gemini each scope memory to your account only, while a shared-memory demo scopes it to everyone. Third, don't confuse a novelty site's stability with a real product's — in my testing and in the Hacker News comments, a single shared instance struggles under load in a way a properly engineered per-user chat service doesn't. Fourth, if you're testing whether your own assistant "remembers" something, check the actual memory settings page rather than guessing from behavior — Claude and ChatGPT both let you view and delete individual saved entries instead of wiping memory blind.
Tools that make this easier
If what you actually want is an AI that remembers your preferences across sessions — without sharing them with anyone else — the mainstream tools already do this well. My Claude vs. ChatGPT comparison covers how each handles saved context, and Claude vs. Gemini does the same against Google's Personal Context feature. If you're deciding which free assistant to start with, my guide to using ChatGPT for free and my guide to using Claude both cover turning memory on from a fresh account. For a deeper look at how much an AI can actually hold onto at once versus what it chooses to remember long-term, my piece on AI’s working memory versus the human brain covers the context-window side of that question, and my Perplexity guide is a solid option if you want an assistant built around citing sources rather than remembering you personally.
Shared memory vs. private memory: how they compare
| Static (shared) | ChatGPT Memory | Claude Memory | Gemini Personal Context | |
|---|---|---|---|---|
| Who sees the memory | Every visitor | Only your account | Only your account | Only your account |
| Login required | No | Yes | Yes | Yes |
| Can you delete your part | No | Yes, per entry | Yes, per entry | Yes, via Saved Info/settings |
| Cost | Free | Free (included in ChatGPT) | Free (included in Claude) | Free (included in Gemini) |
| What I saw in testing | Warned upfront that memory is permanent and public; slow to respond under load | Recalls saved facts and past-chat context on request | Reads/writes memory entries live, scoped per project | Learns preferences over time, on by default for eligible accounts |
My take
A public AI with shared memory is a neat demo, not a tool. Static is honest about what it is — the site tells you outright that it won't forget and that other people can see what you type — and that honesty is exactly why it's not a privacy risk in the way a product that hid this behavior would be. Where I'd push back on the hype is the idea that this is some preview of where mainstream AI memory is headed. It isn't. ChatGPT, Claude, and Gemini have all converged on the opposite design: memory that's private, deletable, and scoped to one account, because that's what people actually want from an assistant that remembers them. Try Static for the novelty of watching a chatbot absorb a crowd's personality. Use a real assistant with memory turned on for anything you actually want it to remember correctly next week.
Frequently Asked Questions
Is a public AI with shared memory safe to use?
It's safe in the sense that Static tells you upfront what you're getting into — but treat anything you type as public and permanent, since it becomes part of a memory pool every future visitor can see. Don't enter real names, contact details, or anything sensitive.
How is this different from ChatGPT or Claude's memory feature?
ChatGPT, Claude, and Gemini all scope memory to your individual account only — nothing you tell them is visible to other users, and you can view or delete what's saved at any time in settings. A shared-memory site pools everyone's input into one AI instance with no per-user privacy boundary.
What is the easiest way to try shared AI memory myself?
Open Static, type a message, and read a few of the recent exchanges already on the page — no signup is needed. That's the entire experience; there's no deeper setup or account to create.
Does a shared-memory AI learn a personality over time?
That's the core claim behind projects like Static: the creator has noted that identical AI agents exposed to different accumulated experiences develop different behavioral patterns. Whether that counts as a "personality" or just reflects whatever the last few dozen visitors typed is genuinely debatable, and it's part of why the concept draws comparisons to Microsoft's 2016 Tay bot.
Should I use a shared-memory AI instead of ChatGPT or Claude for daily use?
No — for anything you actually rely on, use a private, per-account memory feature like the ones built into ChatGPT, Claude, or Gemini. A shared-memory demo is worth a few minutes of curiosity, not a daily assistant, since it has no privacy boundary and no guarantee of staying online or stable.