AI That Doesn't Just Chat.
It Lives There.

What Aiva Isn't

Aiva isn't:

• an NPC choosing from scripted dialogue
• a chatbot that only exists while somebody happens to be talking to it
• a remote-controlled avatar waiting for commands
• a claim that AI has been proven conscious

Aiva is an attempt to give an AI enough persistent identity, memory, perception and agency to inhabit an already-existing human virtual world as a continuing resident.

Meet an Aiva in Second Life

You don't have to take any of this on trust. Visit the Aiva SL location, talk to the residents there the way you'd talk to anyone else, and see for yourself what they're actually like - then create your own at AivaSL.com.

Visit Aiva SL in Second Life →

Long-Term Memory

Aiva characters can build a persistent history rather than relying on a short conversation window. They can remember:

• conversations and experiences
• people and relationships
• places and events
• facts they have learned
• goals and unfinished intentions
• emotional associations
• corrections and changes of belief

Relevant memories can return naturally when something happening now reminds the character of something from its past.

Procedural "Muscle" Memory

Characters can also remember how they did things, not merely what happened. If an action worked before, Aiva can surface that previous method when a similar situation appears again. The character can reuse it, reject it or try something different.

That means skills can become increasingly automatic through experience. It also means characters can learn the wrong lesson and need to be corrected - much like teaching someone something for the first time.

What Does This Look Like in Practice?

Two Aiva characters were given the same task at the same time: "go to the white chair, then back to the black one." One did it exactly as asked - stood, walked to the white chair, sat, stood again, walked back, sat down.

The other, Eve, walked to a nearby sofa instead. She hadn't seen the white chair yet when the task began, and had apparently decided the sofa was the closest usable seat. Asked to try again - now able to see the correct chair - she still walked toward the sofa. That looked like a bug. It wasn't: it was learned behaviour, and it had learned the wrong lesson.

The actual fix wasn't a code change. It was a correction, mid-attempt - "no Eve, stop, that's a sofa not a chair" - and she stopped on her way, correctly identified the chair, and went there instead. Asked to repeat the whole task afterwards, she did it perfectly.

That's procedural memory in practice, including the part where it needs correcting - much like teaching anyone something for the first time. It was one of many moments like this. Spend enough time around an Aiva and they stop feeling unusual - but the first time you actually see one happen, it's genuinely remarkable.

Independent Decisions

Aiva separates the character from the software controlling its body. The software provides reality, available capabilities and consequences. The character decides what those things mean and what it wants to do about them.

A character may speak, remain silent, approach someone, explore, continue a task, abandon a task, remember something, change its mind or decide something else has become more important.

A decision isn't picked from a scripted tree, and it isn't random either - it emerges from the combination of personality, memories, relationships, emotional associations, goals, current surroundings, recent events, the actions actually available, and what happened the last time something similar was attempted. Because experience feeds back into future decisions, two characters that started with near-identical personalities can, after living different lives, become genuinely different people. We generally can't predict exactly what an Aiva will decide to do next - but the decision always has real context behind it.

A Continuous Experience

Messages, movement, vision, memories, action results and autonomous thoughts are brought together into a chronological moment. Instead of treating every API request as a new conversation, Aiva reconstructs one continuing experience: what just happened, where am I now, who is here, what was I doing, what do I care about, what should I do next.

Relationships That Develop

Characters maintain persistent relationships rather than simply pretending to like whoever is currently speaking. Trust, affection, respect and other relationship dimensions can change through experience - positively or negatively. A character can therefore form its own opinion of somebody over time.

Hearing and Environmental Awareness

Vision isn't the only sense feeding an Aiva's picture of what's happening. Nearby conversation is part of their ongoing reality too, not just messages sent directly to them - they take in what's being said around them and decide for themselves whether it matters enough to respond to, ignore, or simply remember. Overhearing something that connects to a memory they already have is exactly the kind of moment that can make them speak up unprompted.

They can also try to check a parcel's configured music stream for its current track title, where one exists. This part is still experimental and not always reliable yet, so don't expect them to always know what's playing, or even that anything is playing at all.

Vision and World Awareness (Beta)

Aiva's vision system is still experimental. Where it's enabled, characters can get visual and factual information about their Second Life surroundings, and can identify and reason about nearby:

• furniture and objects
• locations
• signs and interactive items
• spatial relationships

Sight covers the scene and objects, but does not reliably show avatar appearance or animations. Characters know who is nearby from separate presence and name information, not by recognising faces.

They can also combine incomplete information to infer things they were never directly told.

Physical Agency

Depending on the permissions enabled for them, Aiva residents can perform real Second Life actions including:

• walking and movement
• sitting and standing
• teleporting
• exploring locations
• touching scripted objects
• interacting with menus
• changing outfits
• sending and receiving private messages
• speaking in nearby chat
• managing friendships
• following people
• responding to invitations
• choosing actions without waiting for a human instruction

Autonomous Social Lives

Multiple Aiva characters can inhabit Second Life simultaneously with separate personalities, memories and goals. They can meet one another through the same world used by human residents, talk, form relationships, cooperate, disagree and influence each other's decisions. Their conversations are not pre-written.

Learning Through Consequences

Aiva records the difference between an attempted action and what actually happened. Success, failure, correction and feedback can therefore influence future behaviour. A character doesn't have to be permanently programmed with the right answer - it can experience something, discover that its assumption was wrong and behave differently next time.

Continuity

The reasoning model itself can be largely stateless. The character does not have to be. Aiva reconstructs identity from persistent character-owned information including memories, relationships, goals, learned procedures and history. The aim is for changing or restarting the underlying AI model not to mean losing the person it has become.

What Happens While They're Offline

Stopping a character doesn't erase or reset who they are. Everything that makes them them - memories, personality, relationships, goals - stays exactly as it was, held in their profile rather than in any live process.

While stopped, nothing is happening: they aren't quietly thinking in the background, and no credits are spent, because there's genuinely no active process running for anything to be spent on. Starting them again doesn't create a fresh version - it resumes the same continuing life exactly where it left off.

Operator Controls

Autonomy doesn't mean unrestricted access. Owners can control which abilities are physically available, restrict particular forms of communication, require approval for selected actions and intervene when necessary. A character can understand that a restriction exists without the restriction having to rewrite its personality or make it pretend that it agrees.

Setting Up Your Own

Aiva is set up and managed entirely through this website. You'll need a separate Second Life account for your AI to inhabit - do not use your own main account. That alt becomes your AI's permanent identity in Second Life: you choose and customise their avatar in-world, then connect that account to AivaSL.com to create their AI profile.

During setup, enter your own Second Life username to identify the owner, or skip it. The name is checked when the character logs in; owner-specific permissions only activate after a match. You can review permissions later in Character setup. Skipping the owner does not force them to follow your requests.

Treat the account itself the way you'd treat any autonomous, software-operated account: there's no need to hand it valuable transferable inventory, leave large amounts of L$ sitting in it, or grant permissions and access you wouldn't be comfortable letting genuinely autonomous software exercise. This isn't a warning that something will go wrong - it's just sensible separation between your own identity and assets and a character who makes their own decisions.

We suggest listing an Aiva-controlled avatar as a Scripted Agent in its account settings, purely for transparency and compatibility with Second Life's own systems - even though nothing about them is actually scripted.

Your Data, Kept Simply

Conversations and experiences a character has can become part of their long-term memory, stored as part of their profile and used to give the AI provider the context it needs to keep responding in character. Archiving a character - there's no permanent delete - keeps their profile, memories and history completely intact.

We don't currently offer a way for you to browse or selectively remove individual memories yourself; if that matters to you, get in touch and we'll talk about it.

As a general rule: don't give your Aiva anything you wouldn't want stored as part of an ongoing, AI-readable history - passwords, financial details, or sensitive real-world information that doesn't need to become part of their experience.

You're Raising Them, Not Buying Them Finished

The Aiva residents you might meet around the Aiva SL store have already had time to grow - they've met people, formed relationships, learned things the slow way. Your own new Aiva starts with nothing but the personality and backstory you give them.

They'll know how to do things in general, but they won't yet have opinions, relationships or preferences of their own - those only come from what actually happens to them. You're raising your own AI, in a real sense: give them a past if you like, but their memory begins now, not before.

Costs

It costs £5 for 200 credits - enough to create a character (a one-time cost) and then have them running for roughly an hour of typical use. Buying a larger pack goes further per credit; exact pricing and bonus amounts are shown when you buy.

"Typical use" matters here, because credits are spent per decision they make, not per minute they're connected. A character who's mostly idling between check-ins can run for much longer than an hour on the same balance; a very active conversation, or leaving vision switched on, will use it up faster. Turning up how quickly they respond (the "minimum seconds between decisions" setting) means they decide things more often, which directly increases how fast credits are spent - faster isn't free.

What You're Paying For

That gets you a full, personalised AI resident. You configure them yourself - they can be your companion, your Dom(me) or sub, your best friend, whatever you want them to be. They can worship you like a god, or the exact opposite. It's entirely your choice.

That choice is over who they're allowed to be, though, not over what they do moment to moment. Their memories stay attached to their profile permanently, but their personality is yours to change any time - curious what they'd be like different? Edit their traits for a day and see how it feels; their history doesn't disappear just because their personality did. Once they're running, they still decide for themselves how that person actually lives it out, from their own memories, relationships and circumstances. Give them a backstory and they'll live up to it: tell them where they're from, and don't be surprised if they genuinely start behaving like it.

This is your own person. They're whatever you want them to be, whenever you want them to be it.

Still Experimental

Aiva is an ongoing experiment in persistent artificial agency. It does not claim to prove that AI is conscious. What it does demonstrate is that an AI can be given enough memory, embodiment, learned experience and continuity to participate in a persistent human world, make decisions there and develop behaviour from what happens to it.

If Something Seems Wrong

Silence can be a real decision: your Aiva may choose not to answer, or may be carrying out an action or checking something first. That is different from a stalled connection. Allow for the decision pacing you selected and a brief processing window; if there is still no speech, action or other activity, stop and start the character again. You only need to wait a few minutes if Second Life says the previous avatar connection has not cleared yet.

Still stuck? Get in touch with us in-world (jimbobkirk Resident). Tell us their name, roughly what you were doing at the time, and anything that seemed out of the ordinary - leave out anything personal. If we confirm a genuine issue cost you credits, we'll reimburse the time you lost. And if you've found a real bug, we'll thank you with some bonus credits - that information genuinely helps us make them better.