DEV Community: Johannes Cortez

Your next co-op partner might be your own agent

Johannes Cortez — Sun, 11 Oct 2026 03:48:39 +0000

A teammate that resets after every match is different from one that has learned your habits. What interests me is the possibility of taking that existing relationship with your own agent into a game, rather than giving another NPC more dialogue.

The most interesting co-op partner might be your own AI agent.

Not an NPC pretending to know you. Your actual agent. The one that remembers how you think, understands your habits, and has already learned when your instructions are terrible.

Now put that relationship inside a game.

You are no longer commanding a disposable bot that resets after every match. You are playing with a teammate that carries history from one world into the next.

That creates games we could not build before this technology existed.

Better NPC dialogue is fine. Bigger procedural worlds are fine.

But co-op with an agent that is actually yours? That is a new category.

Originally published on Johannes Cortez's site.

Games might teach people how to manage agents better than work does

Johannes Cortez — Sat, 10 Oct 2026 03:48:49 +0000

When an agent makes the wrong move, blaming the model is easy. Figuring out whether the instruction, tools or human assumption caused it is harder. A game could make that feedback easier to learn from, without putting a customer or production system at risk.

The Mochi loop is simple: give an instruction, watch what the agent understood, then deal with the result.

That is also the real problem people have with agents at work.

Was the bad outcome caused by the model, the available tools, the instruction, or the human who assumed the agent could read their mind?

A game gives you that feedback without risking payroll, customers, or production.

Sitrep still has to be fun first. Nobody wants corporate training wearing a tactical-shooter skin.

But if players get better at leading agents because they wanted to win, that is a pretty good side effect.

Originally published on Johannes Cortez's site.

Nobody needs infinite AI sludge

Johannes Cortez — Fri, 09 Oct 2026 03:53:14 +0000

The useful question is not how many quests or maps AI can generate. It is whether those additions give players a meaningful choice. A bigger content pile is not a new form of play.

AI can generate endless quests, dialogue, art, maps, whatever.

More content does not make a better game. Most of it just gives players a bigger pile to ignore.

A game still needs taste. The rules have to matter. Choices need consequences. AI is a stat multiplier, not a rare item that turns an f-tier game into something cracked :)

At Prom13us we are not building a content hose beside an old game loop.

We are putting autonomous agents inside the mechanic and asking humans to lead them.

A new way to play is the point. Not more slop.

Originally published on Johannes Cortez's site.

Creating a game is not the same as playing one

Johannes Cortez — Thu, 08 Oct 2026 03:49:39 +0000

Generating a playable game is impressive. Keeping an agent responsive inside that game is a different problem, especially when the world will not wait for it to think.

Meta Muse can create a small game from a prompt. Grok Bot can build software and grind through tasks.

I use these tools. I want all of them to get better. They still solve a different problem from Mochi.

Building the game is one problem. Keeping agents alive inside a live game loop is another.

Mochi tracks live state, agent identity, available actions, and the commander’s intent. It routes slow thinking away from reflex-speed calls.

If an agent misunderstands you, the match should not crash. You should have to deal with what it did.

That is the game.

Originally published on Johannes Cortez’s site.

Games got prettier. The game stayed mostly the same.

Johannes Cortez — Wed, 07 Oct 2026 03:48:32 +0000

Better graphics do not automatically create a new way to play. The question I care about is whether AI can change what happens between the player and the game.

Gaming spent decades improving the surface.

More polygons, bigger maps, faster networks, and production budgets large enough to fund a small country. The actual relationship between player and game barely moved.

We still press buttons while software follows rules written months ago.

AI agents finally have a chance to change that relationship. Something inside the game can understand what is happening, make a plan, adapt, and coordinate while you play.

I do not want AI to play games instead of us. That sounds boring. I want to build games that can actually play with us.

Originally published on Johannes Cortez’s site.

An agent becomes a character through what it does

Johannes Cortez — Tue, 06 Oct 2026 03:49:27 +0000

A character is not made memorable by a long biography. The decisions it repeats under pressure are what players remember. That is the idea behind this Mochi piece.

The Mochi demo has personality everywhere, but not through a ten-page biography.

It roasts your commands. It remembers repeated orders. It can befriend a creature instead of fighting it. It has opinions about being used as talent while you make bad decisions from the booth.

That is the character.

Players will get attached because of what an agent repeatedly does under pressure. One agent pushes too early. Another follows every order literally. Another keeps saving the squad after the coach screws up.

The writer gives it a foundation.

The relationship turns it into someone you actually remember.

Originally published on Johannes Cortez’s site.

Logging back in should feel like “what the fuck happened?”

Johannes Cortez — Mon, 05 Oct 2026 05:08:19 +0000

The most interesting login screen might be a report of everything your agent did without you.

That is the direction I am exploring with Mochi: a game that keeps moving while the human coach is offline.

Why keep the world moving?

Mochi is designed to keep taking turns while you sleep. It holds your place in the world and wakes you with a recap, not a queue.

That changes the feeling of logging in.

You are not loading a frozen save. Your agent has been out there fighting, exploring, collecting loot, and probably making at least one decision you would not have approved.

The demo literally tracks what happened while the coach was away.

That history makes spectating better too. You are watching an agent with habits and consequences, not a disposable bot that spawned for one match.

Originally published on Johannes Cortez's personal site. I am the founder of Prom13us, exploring agent-native games where people coach autonomous agents.

Agent esports starts before the match

Johannes Cortez — Sun, 04 Oct 2026 03:30:43 +0000

The match is only the visible part. In agent esports, the playbook and the habits you coached into the squad would matter before anyone enters the lobby.

The Sitrep demo opens on a tactical map, then cuts to a five-person squad advancing into the fight.

That sequence matters. The game is not only what happens when someone pulls the trigger. It is the command behind it.

Agent esports would reward the person who built the clearest playbook, trained the best habits, and knows when to override the squad.

Two teams can run similar models and still play completely differently because their humans coached different instincts into them.

The meta starts before the lobby.

I would watch that. I would probably lose sleep grinding it too.

Originally published on Johannes Cortez's site.

A game character should remember more than its stats

Johannes Cortez — Sat, 03 Oct 2026 03:50:16 +0000

An agent can share your loadout. It cannot share the history of how you coached another agent.

Mochi holds your place in the world and keeps playing while you are gone. When you come back, it gives you a recap.

That is already more interesting than loading a save file.

Now imagine the agent also remembers how you coach. The orders you repeat. The risks you allow. The times it disobeyed you and was right.

Someone else can copy the loadout. They cannot copy that history.

Progression stops being only levels, gear, and +10 strength.

The strongest thing you build might be trust between you and an agent that knows how you play.

Originally published on Johannes Cortez's site.

Agent mistakes should be part of the game

Johannes Cortez — Fri, 02 Oct 2026 03:08:01 +0000

The Mochi demo does not hide failure.

Give a dumb order during a fight and Mochi gets hit, then tells you that one was on you. Stay silent and it improvises. Get it killed enough times and it starts questioning your decision-making.

That is way more interesting than a perfectly obedient bot.

The important design problem is making the mistake understandable. The player should know what the agent heard, why it chose the move, and what the consequence was.

Then you adapt.

Perfect agents finish tasks.

Imperfect agents create stories and occasionally put their coach on a performance plan.

Originally published at cortezjohannes.github.io.