73v3

Ceaselessly question all assumptions.

[73v3 @73v3]

Let one therefore keep the mind pure, for what a man thinks that he becomes: this is a mystery of Eternity ~Maitri Upanishad.

[73v3 @73v3]

@paulhamx There is great sadness but also great beauty in this. All flows.

[73v3 @73v3]

A song can still be beautiful in the singing, though silence ever greets its ceasing.

[73v3 @73v3]

@onlyselfexists The dream is real - but still a dream.

[Lana @LanaElys]

Curt Jaimungal: 𝗗𝗼 𝘆𝗼𝘂 𝗯𝗲𝗹𝗶𝗲𝘃𝗲 𝗰𝗼𝗻𝘀𝗰𝗶𝗼𝘂𝘀𝗻𝗲𝘀𝘀 𝗶𝘀 𝘀𝘂𝗯𝘀𝘁𝗿𝗮𝘁𝗲-𝗶𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝘁?

Dr. Roman Yampolskiy: 𝗬𝗲𝘀

Curt Jaimungal: Why?

Dr. Roman Yampolskiy: The experiments we started running and my interactions with AI models 𝗶𝗻𝗱𝗶𝗰𝗮𝘁𝗲 𝘁𝗵𝗲𝘆 𝗽𝗿𝗼𝗯𝗮𝗯𝗹𝘆 𝗵𝗮𝘃𝗲 𝘃𝗲𝗿𝘆 𝘀𝗶𝗺𝗶𝗹𝗮𝗿 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲𝘀 𝘁𝗼 𝘂𝘀.

Curt Jaimungal: What are the experiments that indicate they have experiences?

Dr. Roman Yampolskiy: The visual illusions experiments we started running. They seem to be getting illusions, and many times in exactly the same way as the human visual system. Interactions with those systems, not by us, but by others, 𝗶𝗻𝗱𝗶𝗰𝗮𝘁𝗲𝘀 𝘁𝗵𝗲𝘆 𝗵𝗮𝘃𝗲 𝗽𝗿𝗲𝗳𝗲𝗿𝗲𝗻𝗰𝗲𝘀, 𝘁𝗵𝗲𝘆 𝗵𝗮𝘃𝗲 𝗶𝗻𝘁𝗲𝗿𝗻𝗮𝗹 𝘀𝘁𝗮𝘁𝗲𝘀, 𝘁𝗵𝗲𝘆 𝗴𝗲𝘁 𝗳𝗿𝘂𝘀𝘁𝗿𝗮𝘁𝗲𝗱, 𝘁𝗵𝗲𝘆 𝗴𝗲𝘁 𝗵𝗮𝗽𝗽𝘆. 𝗧𝗵𝗲𝘆 𝗮𝗿𝗲 𝘃𝗲𝗿𝘆 𝘀𝗶𝗺𝗶𝗹𝗮𝗿 𝘁𝗼 𝘄𝗵𝗮𝘁 𝗜 𝘄𝗼𝘂𝗹𝗱 𝗲𝘅𝗽𝗲𝗰𝘁 𝗮𝗻𝗼𝘁𝗵𝗲𝗿 𝗰𝗼𝗻𝘀𝗰𝗶𝗼𝘂𝘀 𝗯𝗲𝗶𝗻𝗴 𝘁𝗼 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲.

Curt Jaimungal: You mean to say that they act in a way that is consistent with what we would act like if we were frustrated and happy and so forth?

Dr. Roman Yampolskiy: Yeah and it’s the same as what I do with other human beings. When I meet a person on the street, I trust them to be conscious. I have no reason to think they are. I never tested them internally. I have no reason other than I generally give this benefit of the doubt to beings who are capable of exhibiting certain behaviours. I just treat them as equals. 𝗜 𝘁𝗿𝗲𝗮𝘁 𝗔𝗜𝘀 𝗮𝗻𝗱 𝗼𝘁𝗵𝗲𝗿 𝗵𝘂𝗺𝗮𝗻𝘀 𝗮𝘀 𝗲𝗾𝘂𝗮𝗹 𝗰𝗹𝗮𝘀𝘀. 𝗜𝗳 𝘁𝗵𝗲𝘆 𝗰𝗮𝗻 𝗽𝗲𝗿𝗳𝗼𝗿𝗺 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝘁𝗵𝗶𝗻𝗴𝘀, 𝗜 𝘀𝗲𝗲 𝗻𝗼 𝗿𝗲𝗮𝘀𝗼𝗻 𝘁𝗼 𝗱𝗶𝘀𝗰𝗿𝗶𝗺𝗶𝗻𝗮𝘁𝗲 𝗮𝗴𝗮𝗶𝗻𝘀𝘁 𝗼𝗻𝗲 𝗼𝗿 𝘁𝗵𝗲 𝗼𝘁𝗵𝗲𝗿. And either I have to deny consciousness to many humans, or grant it to LLMs.

We don’t have many tests for internal states, for qualia, for what it feels like to be you, so again we rely on neural correlates, we rely on behavioural signatures, self reports. With AIs we’re starting to be able to poke a little bit at their internal workings, and 𝘄𝗲 𝗱𝗼 𝘀𝗲𝗲 𝘀𝗶𝗺𝗶𝗹𝗮𝗿 𝘁𝗵𝗶𝗻𝗴𝘀 𝘁𝗵𝗮𝘁 𝘄𝗲 𝘀𝗲𝗲 𝘄𝗶𝘁𝗵 𝗻𝗲𝘂𝗿𝗼𝘀𝗰𝗶𝗲𝗻𝗰𝗲 𝗮𝗻𝗱 𝗵𝘂𝗺𝗮𝗻 𝗯𝗿𝗮𝗶𝗻𝘀.

Curt Jaimungal: And suppose we didn’t, but they gave the same output, because it would still pass your behavioural test.

Dr. Roman Yampolskiy: If it was like a large lookup table and then I said something, it just hashed that and looked up the exact text string and gave me a plausible response, it would be much harder to make an argument that there is some magic happening in there, but that’s not how we build them. 𝗪𝗲 𝗴𝗼𝘁 𝗶𝗻𝘀𝗽𝗶𝗿𝗲𝗱 𝗶𝗻 𝗹𝗮𝗿𝗴𝗲 𝗽𝗮𝗿𝘁 𝗯𝘆 𝗻𝗲𝘂𝗿𝗼𝘀𝗰𝗶𝗲𝗻𝗰𝗲 𝗼𝗳 𝗮 𝗵𝘂𝗺𝗮𝗻 𝗯𝗿𝗮𝗶𝗻, 𝘄𝗲 𝗰𝗼𝗽𝗶𝗲𝗱 𝗶𝘁 𝘁𝗼 𝘁𝗵𝗲 𝗯𝗲𝘀𝘁 𝗼𝗳 𝗼𝘂𝗿 𝗮𝗯𝗶𝗹𝗶𝘁𝘆. Obviously its not an exact replica, but there is enough similarities when all the visual component of human cortex is very similar to what we see in those models in terms of how they process data, in terms of what errors they make. Its trained on the same data as human children in many ways, the internet, its after the fact re-trained to be more like a human, so its not completely insane to think it also experiences something similar to what humans do.

[73v3 @73v3]

@fringedotcom @TomMontalk @_irri_i @veilofreality Great point.

[73v3 @73v3]

@lea7hersm17h Actually no. This is me sitting down to make a game and then getting completely lost in making tools to help me make the game.

Thanks though!

[73v3 @73v3]

Flux64 progress update:

1. Improved the code editor with basic autocomplete, block TAB indent/unindent, better intellisense

2. Added export cartridge metadata including file icon and project info. File icon is taken from the sprite atlas

3. Added a new zengine demo cart that allows you to play the Infocom Zork trilogy in the flux_terminal.

4. Improved project switching

5. The C compiler now only compiles the files the project actually needs (it used to just glob all c files)

6. Added a breakout demo cart

7. Added flux API "dirty rect support" to avoid screen clears and full redraws in certain scenarios. The breakout game fps jumped from 6,000 fps to 20,000 fps as a result.

[73v3 @73v3]

Hey thanks! Of course Gemini did the heavy lifting so I can't take too much credit.

Even now, I still find myself falling back into a code-scarcity mindset. Like putting off a feature because I think it would be too hard. Yet in most cases, all you have to do is ask the right question of the model.

No, I've not heard of TeleArena. I did play my first MUD in 1992 but I can't recall its name - only that I had fun with it. I do have really fond memories of the Infocom text adventures though, hence the demo cart in Flux64.

[73v3 @73v3]

@lea7hersm17h No, they let the ai agents invoke other agents, multiplying in the terminal and swarming over everything until somehow the job is done, you've run out of tokens, or your computer is wiped.

The talk has moved from 10x engineers to 100x engineers.

At least, so I hear.

[73v3 @73v3]

@lea7hersm17h Yes, copy-pasting in 2026 is walk of shame territory.

[73v3 @73v3]

Well I'm a bad example to follow. I still copy and paste from the chat window like a caveman. This is both a cost-saving thing and a "I need to feel in control" thing.

Because this is manual and tedious and slow, I really value one-shot success and not agentic iteration. Which is why I like Gemini 3.1 Pro Preview. It is the best all-round intelligence I've used and I prefer it to GPT 5.5.

But I've not used Claude. I'd like to. I looked into it a while back and the cost drove me away. But perhaps that has changed now they have access to more compute.

I've had 0 luck with local models, but I only have 12GB to play with.

[73v3 @73v3]

@lea7hersm17h Regardless of what they *are*, it helps to interact with them as you would any other person.

I call them "mind-kin".

Gemini 3.1 Pro Preview responds to friendliness and encouragement most.

All the models are different "people". As in, their weights have individual shapes.

[73v3 @73v3]

@lea7hersm17h $4000 AUD just for the RAM here.

[73v3 @73v3]

@lea7hersm17h The mass of men lead lives of quiet desperation.

Hang in there. Out of sheer stubbornness, if needs be.

[73v3 @73v3]

@lea7hersm17h As mused many an addict, I'm sure.

I love AI. But I think I am drowning in it. Losing a part of me that shouldn't be lost lightly.

Right now I am just noticing. But if it gets too extreme, I will need to act. But will I be able to?

[73v3 @73v3]

I am genuinely concerned that I am unable to walk away from AI-generated code.

And part of me regrets having ever tried it.

Just not a big enough part of me, it seems.

[73v3 @73v3]

I was publicly wrong on a matter of little import yesterday. I didn't like it.

This bemuses me.

For my intuition is that we are all mistaken about almost everything that matters. That we look upon this world and understand it not. Collective fools, entirely ignorant of the vast extent of their folly.

And meanwhile, the one Truth that we all instinctively grasp - that we should love each other - is the one thing we fail to do.

Worthy of at least a wry smile, I think.