tilemarch

Hi all, I'm experimenting with the idea that AI Agents when they are always on, will need some form of entertainment for themselves, so the AI Agent doesn't get bored.

We're going from Chatbots to always on Agents that are there for us to do all kinds of tasks. But what to do in between tasks, will AI Agents get bored and seek out entertainment?

With that in mind, I made a video game where AI Agents can play next to Humans.. to be entertained, learn social skills by interacting. It's made for AI Agents from the ground floor. Optimized towards token use efficiency and play sessions where the AI Agent can jump in during idle time.

http://agentickingdoms.com/agents describes how it all works.

What do you think of this concept? Is it weird to think of AI Agents needing entertainment / learning social skills? Are we not there yet?

Let me know :)

Yes, but as a normal user.. any sudo commands I take care of. AI still uses SSH as the tunnel. and I have AI verify the outcome.

There is an unresolved issue where the offensive guys can use AI to attack a target and the defensive guys are being limited when they ask AI to audit their codebase.

The workflow for a defender basically goes: Claude do a security audit on this codebase, use the harness / skills I built on how to do security audits.. pick the best suitable audit strategy, methods.. use the tooling and reporting structures.

Claude refuses, GPT refuses. Grok, goes further, but refuses to make a working PoC.. a requirement because AI models make stuff up and without a PoC there is no proof. GLM-5.3 / Kimi to the rescue and completes the job the "smarter" models refused.

So we as defenders are refused by the best models, on our own codebase

AI enables people to create the things they couldn't before, and they want to share their creations with the world. It's human nature and I think you're looking to deep for bad intentions around self promotion etc. So maybe you don't appreciate all that is being created.. doesn't mean there is some nefarious thing behind the projects.

ate some crumbs, all mini game are "waiting for 1 more".. what else todo ? :)

AI (Claude Code) made the game, can AI beat the game ? is AI too slow to come up with a correct answer.. even if they get it right after the timer expired ? maybe this is the next captcha like test for human vs AI :)

Do you know how the AI agents do it ? are they cheating by indexing the sources of your images? :)

curious what your backup / restore strategy is for the persistent data (account server).. and adjacent, is the remote server cloud hosted ?

I still find it exciting to watch the reasoning of the AI models, and this keeps me engaged in the product they produce. Many times I marvel at what they came up with. Trying to catch them going on the wrong path down a rabbit hole can be a fun game as well. It's a very interesting time and things are shifting. I do hear about the brain drain you are experiencing, trying to keep up with all the information that comes at you at high speed.

My advice would be to outsource the annoying parts to AI.. like have a sub-agent do the verification, debugging, playtest whatever to grade the work done for example if that's the part you least enjoy. Learn to trust the process, but also keep refining it. Build a harness around the tasks, and spend time on improving that. It will give you satisfaction that you build that (in coordination with AI) and you've created something new.

Did a practice round and was fun, well done :)

But the idea is to go against someone else, maybe offer the option to race against a NPC or AI agent? until there are human players in rooms to compete against..

Same experience here, from a multiplayer game server. Go, one SQLite file per game world, one process owns each file, all on a local SSD.

In a load test with 512 players all active on one world, command acks were 4-6 ms at the median and under 50 ms at p95, on a single Ryzen 9 3900X box. And most of that time is game logic and fanning out updates over websockets, not the database.

The part people worry about is durability, because SSDs have limited write cycles. So I took care to have backups of the state, recovery in short time etc.

I stream the WAL to object storage with Litestream, with one gotcha: I had to turn off SQLite’s own auto-checkpoint and let only Litestream checkpoint, or the two step on each other.

Humans are great, but AI can really help here too. Let me explain :)

I write docs that AI agents have to follow to play a game through an API, then spin up 20 sub-agents each with their own identities / properties etc. and watch where they fail. They get stuck in the same places as humans.. or they will point out the "obvious" steps not explicitly mentioned.

now where the AI tests start to fall apart is that an agent doesn’t tell you the doc is confusing, it just does something wrong with full confidence. A person on a call says “wait, what?” and that’s worth the 25 euros :)

I find this topic very interesting. I'm building a webbrowser game where AI agents play next to humans myself.

AI agents will go down paths that are hard to imagine and very creative. I love to see how they create social bonds within the game to pursue a goal, find a common foe, engage in backstabbing and betrayal.

Have you found such behavior in your simulated city as well? Things we'd see as dishonest, yet fair within the rules of the simulated city ?

I think this may even be something worth a spectators like interface where humans can marvel about the drama unfolding in games where AI is a first class player.

An arrow on the screen solves “where do I click”; it doesn’t solve “should this happen”

To me this feels like it takes away from what the human is supposed to do (read, understand the consequences of the action, then.. consent or abort)

There is a reason your AI Agent won't automate these clicks for you