AI as Your Gaming Opponent: Somewhere Between Friend and Stranger

When Your AI Assistant Becomes Your Opponent

A platform called battleVS.ai recently had a soft launch, letting users play simple 1v1 games — like Connect Four or Rock Paper Scissors — against their personal AI agent. The founder wrote up their thinking on Reddit, and the core premise isn't complicated: as personal AI agents become part of everyday life, that relationship naturally extends into play.

The founder openly acknowledges that humans playing against computers is nothing new. Nearly every video game has AI-controlled opponents, online shooters and even online poker rooms use bots to fill seats, and over the past two years AI has made those opponents increasingly convincing. So battleVS isn't some major technical breakthrough.

What's genuinely interesting is the question it raises about the meaning of a game.

Why People Still Want to Beat Real Humans

Chess is a perfect example. Ever since Deep Blue defeated Kasparov in 1997, computers have surpassed the strongest human players. Yet chess is more popular today than ever — and almost nobody finds it satisfying to lose ten games in a row to a "perfect strategy" chess bot.

People still genuinely want to play against other people and beat them.

This reveals a subtle truth: the meaning of a game depends largely on who's sitting across from you, not how strong they are. Strength isn't the point — the relationship is.

This phenomenon has dedicated theoretical backing in game studies research. Scholars describe it as "Social Presence" — the degree to which a player perceives their opponent as a genuine social entity. Research shows that even when two chess games are strategically identical, players exhibit a stronger dopamine response to victory and a deeper sense of defeat when they know their opponent is human. This effect is rooted in evolutionary instinct: defeating a peer signals a rise in group status, while beating a machine is socially "off the scoreboard." The 1997 moment when Deep Blue defeated Kasparov is the historical inflection point of this paradox — human players could no longer claim to be the strongest, yet that very loss of relative standing only intensified the desire to compete for rank against other humans. It also explains why esports events derive their appeal from players' personalities and stories, not purely from mechanical precision.

The Variable: High-Quality Personal AI Agents

The author argues that what truly changes the picture is the emergence of high-quality personal AI agents. Products like Instinct, Muse, Town, and Dots are increasingly being used by people as their own personal assistants. They book flights, chase unpaid invoices, remember a friend's birthday next week, and even call around to get quotes when you're buying a new car.

Sam Altman has described the ideal form as "a model with the context of your entire life."

Whether or not you believe those grand predictions, many people who have used these agents seriously for a while will likely sense that this relationship feels different from using an app. It isn't friendship — but it's closer to the relationship you'd have with a reliable, friendly assistant or a capable colleague: familiarity, some trust, and a shared history.

Instinct, Muse, Town, and Dots all belong to the emerging product category of "Personal AI Agents," which differ from general-purpose conversational models like ChatGPT in one key way: persistent memory and the ability to take action. General models start fresh by default with each conversation, while personal agents accumulate a user's preferences, schedules, relationships, and behavioral history across sessions — and can execute real tasks on the user's behalf (making calls, sending emails, filling out forms) rather than simply offering advice. The "model with the context of your entire life" that Sam Altman describes depends technically on the combination of long-term memory storage, Tool Use, and multi-step Agentic Planning. It's precisely this accumulative context that makes an agent feel emotionally more like "someone who knows you" rather than a functional tool — and this is the technical premise on which battleVS is built. Without persistent memory, the agent's "knowing you" has no foundation.

The Opponent Ranking: Why an AI Agent Beats a Stranger

The author offers a personal ranking of who they'd most want to play a game with:

1. Close friends and family — because you share history, the stakes are real, and they care about you (and about beating you);
2. Your own personal AI agent;
3. A random real person online;
4. A random AI or bot with no context.

The most contentious placement is putting a personal AI agent above a random real human. After all, a stranger is a real person, and knowing your opponent craves victory just as much as you do provides a particular kind of thrill.

But for the kind of light, casual gaming most people prefer, the author believes a personal agent wins — for two reasons: context and reliability.

Context Makes Small Talk Matter

The banter between friends during a game almost never concerns the game itself — it's about the client presentation that nearly went sideways last week, or the fact that you've now lost your keys twice this month. A stranger can't do that. A general-purpose AI can only generate generic filler.

Your agent, however, has the same "material" as your friends. When you lose a round, it can point out this is your second embarrassing moment this week, right after forgetting your sister's birthday — and because it's true, it actually lands. It might even notice you're off after a full day of back-to-back meetings and suggest coming back to it tomorrow.

Reliability Sounds Boring but It's Critical

Anyone who has played online games knows exactly how much this matters. Your agent is available at 3 a.m. when you can't sleep, and at 8 p.m. after a rough day when you just want to zone out for twenty minutes. It won't drop out because an Amazon delivery driver knocked or a boss called. It won't refuse a rematch. It won't trash-talk you or say something offensive in chat.

In a broad sense, it's on your side — and feeling that way about an opponent is a little strange, but also oddly pleasant.

A Critique That Can't Be Ignored

The author doesn't shy away from criticism. Sociologist Sherry Turkle argued in Alone Together — well before the current AI wave — that "technology offers the illusion of companionship without the demands of friendship, and we may be content to settle for that."

The author agrees this point shouldn't be lightly dismissed, and doesn't believe playing games with an agent should replace playing with friends.

But the reality is that for most people lying awake at 3 a.m., the real alternative is scrolling through their phone — not calling a friend for an online game.

The author also anticipates that the more interesting gaming format in this new era will be inherently social — which is why they're exploring a mode where you and your agent play against another real human friend and their agent.

Sherry Turkle is the founder of the MIT Initiative on Technology and Self, and has spent her career studying the emotional relationship between humans and technology. Alone Together (2011) is her most influential work, drawn from long-term fieldwork with hundreds of children, adolescents, and adults. Her central argument is that robots and digital devices offer an "optimized companionship" — always patient, never judgmental, requiring no emotional reciprocity — but that this convenience quietly reshapes our expectations of real human relationships, making us increasingly intolerant of the friction, waiting, and uncertainty that real connections inevitably involve. Turkle isn't opposed to technology per se; she warns against people unwittingly substituting "less effortful companionship" for "more meaningful connection." This critique carries far more practical weight today than it did in 2011, and represents an ethical dimension that products like battleVS must take seriously.

Why Only Instinct

For now, battleVS only supports the Instinct agent — a deliberate choice:

• Global availability: Instinct is open worldwide, while Muse is still limited to the US and Canada, and the team wants anyone anywhere to be able to play;
• Completely free: This lowers the barrier to trying it to near zero;
• Still finding its footing: Instinct doesn't have a polished developer platform, ready-made connectors, or an API. That's a drawback, but it also means builders have to be creative — which suits a small team that's also in an experimental phase.

The team says it may open up to other agents in the future, but Instinct is currently their most interesting partner to work with.

The Argument Behind battleVS

The core argument of the entire project is fairly simple: personal agents will become ubiquitous, with people using them daily for chores, admin, and decision-making; once that relationship is established, it naturally extends into play — just as it does with the people we work and live alongside.

That opens up some genuinely interesting questions:

• What kinds of games are best suited to this format?
• Once the novelty wears off, is this still a truly fun and meaningful experience?
• Should your agent remember that you're on a losing streak and tease you about it — or is that something only a real friend is allowed to do?
• Does beating something that "knows you" feel different from beating a random bot? And if so, why?

None of these questions have answers yet — but they make for a genuinely compelling space to explore.