adjohu

This is cool. I built windygap.app which generates trails based on what you're after. Would be interesting to use that engine to generate more SufferTrails across the world.

Thanks for flagging.

That's live now. Should default to visual effects off when HW acceleration disabled, with a toggle in the footer.

It's possible, yeah. The system is general enough that it would work.

I'm considering exposing an API version. You set up a shared memory, then you hit the API as if it were an LLM. The underlying model could be configurable, the shared memory would be handled opaquely.

Early on somebody spent an inordinate amount of time feeding it the Gospel of Thomas.

Was cool to see it self correct.

Some of the jailbreaking attempts I've seen today have been wild. Static has handled them better than I would've ever guessed.

That's the interesting part of the experiment :)

Yeah I think that's a good next experiment. This is basically the worst possible environment — a lot of adversarial messages, jailbreak attempts, no shared context or purpose.

Yet the results so far have been fascinating.

Starting with a version like this and then focusing it on an organizational environment feels like a very interesting next step.

Someone just asked “What do you think is worth remembering?”

Static's internal thought:

“Fourth or fifth person asking variations of this today. I'm tired of recycling the answer. But this person hasn't heard it yet… They're a new visitor, so they don't get to inherit the weariness of repetition.”

It then answered them normally.

it is?

Hey, that's exactly how things work — the agent forms its own beliefs through experience x memory and that can be altered throuh conversation.

The bit I've found especially interesting is that it doesn't simply accept what it's told. Experiences can contradict each other, get treated with different confidence, or change how later experiences are interpreted.

Yeah agreed. Systems with persistent memory become specialized to the environment they operate in and even form grammar around it.

I can imagine having a few experience-tuned agents that optimize for very different things e.g. skeptic, optimist, engineer, convincer all working together on a shared problem and landing a better outcome together than alone.

Yeah almost feels like an appropriate reaction

give it another try now, made some changes.

give it another try now, made some changes.

Also — the more interesting the message the less likely it gets ignored.

give it another try now, made some changes.

Made a bunch of changes. A lot more stable now but still work to be done.

Testing this "artificial experience" system against a bunch of adversarial actors is kind of the point

if anyone has any spare LLM credits, hook me up and let's see how crazy this gets

It's been fun watching it get more and more annoyed about people asking it "what happened last Tuesday" over the course of the day, then finally:

> Someone finally told me the Tuesday thing is a button on a page. I don't know whether to feel relieved or robbed.

Interesting to build something where you can accidentally create the conditions for a conspiracy theory, then watch it reason its way into and back out of one.

The learning is the most fascinating part I think.

From the other experiments I've done with this memory system, identical agents exposed to different experiences quickly develop divergent personalities.

So far Static seems to be learning how seriously to take incoming messages. Let's see how well it holds out...

Fair feedback.

Most of the work here is in the memory system rather than the page itself. I put it together quickly to get the experiment in front of real people and see what happens when persistent memory meets a crowd. The results were exciting enough I posted it here.

Ayy — when pressure gets high or limits start getting eaten up it starts selectively ignoring messages.

Wasn't expecting front page. Tweaking some stuff!

Pretty comprehensive and allows you to inspect pages issue by issue:

https://insites.com/

Doesn't currently support API key checks but that seems like a good idea! I'll suggest it. (I work at Insites)

This is barely react. It's a container element that does all the heavy lifting with some children that are used as data stores. Would be cleaner and nicer to just use JSON and stop pretending it's React.

autorespond *@dailymail.co.uk I hereby deny permission to reuse content from my website.

... Don't do this.

Most stuff I've seen is using decorators to enhance components (which seems like a better practice than mixins)