0xAyuki

Professional exit liquidity provider

[Elfa AI @elfa_ai]

Introducing Iris.

The first-ever real-time intelligence stack for AI agents.

A new era of context-aware agents starts now.

[0xAyuki @0xayuki]

@j0hnwang Definitely @matrixthesun

[0xAyuki @0xayuki]

@0xShual Extremely beautifully written

[Shual @0xShual]

I've been wrestling with a thought I can't quite articulate cleanly, but I'll try. I know I hardly ever do this nowadays, but beware: this one's a long ass tweet.

There seems to be a curve people travel when they engage seriously with AI. It feels a bit like the market cycle graph, or even someone who tries psychedelics for the first time. I'll try to explain.

The first phase is genuine awe. I've seen non-technical people discover what these tools can do and lose their shit. (@0xkyle__ for example, who the quoted tweet is responding to).
They instantly clear their inboxes, automate their calendars, spin up applications without writing code, and suddenly feel like the interface between intention and execution has collapsed. "I am a god now, the limit is my imagination", and this is actually correct about something: the interface to capability feels fundamentally changed.

The non-technical person isn't wrong that they can now do things that previously required either technical skill or hiring someone with technical skill. That's a real thing. Capability has been democratized and made easily available.

Then there's the technical person's retort (quoted tweet).
"It's a glorified text generator. It has no emergent thinking. The more you understand the architecture, the less impressed you are." These people often point to the failures ("make shitty code", "inability to do novel things", "they're brittle", "ppl are just mass producing slop daily task managers and not meaningful, complex products"). They are also correct. But at a different level of description.

Yes, it's predicting tokens. Yes, it doesn't have persistent memory, can't truly learn within a conversation in the way a human does, makes confident mistakes, and has real architectural limitations. But this (reductionist) critique sometimes smuggles in an assumption that if you can describe the mechanism, you've explained away the phenomenon.

I don't know if these systems think. I'm not sure I know what thinking means precisely enough to answer that. What I notice is that people who are certain it's not thinking seem very confident about what thinking is, and I don't think anyone's earned that confidence yet. We point at our own cognition and say "that, the thing I do"--but we can't actually specify what we're pointing at.

The ceiling question is where I find myself most uncertain. The "AI can't do X" predictions have been wrong often enough that I don't put much stock in confident limits. At the same time, current systems have obvious constraints: no learning within a conversation, errors that feel brittle rather than like honest mistakes, moments where it seems like performance of understanding rather than actual understanding.

And then there's a third position-- the researchers. The experts among experts. The people who are tracking and actively pushing the field forward. They're not dismissive. They're certainly not naive. They're watching the _curve_ and wondering/fearing where it goes. Not what the system is now, but what the curve suggests about what's coming, and whether we have the wisdom to navigate it.

When someone says "AI is just predicting the next token," they're describing the mechanism. When someone says "it built me a working application," they're describing the phenomenon. Not contradictory things. Orthogonal.

Water is "just" hydrogen and oxygen, but that tells you nothing about what water does, how it behaves, why it's wet. The reductive explanation is true but incomplete.
"Just statistics" often gets said like it settles something, when really isn't that where the interesting questions begin? What patterns, matching against what? And why does that yield something that can engage novel problems and hold context across a conversation?

When someone first takes psychedelics, and I mean a proper dose, that experience fundamentally changes their priors. (recent public example is @bryan_johnson) It rearranges something in your model of reality.

You see (not 'understand', but see) that perception is constructed, that the self is more... porous than it feels, that consciousness has textures you didn't know were there. And then you tell someone and you immediately worry how insane (or banal) you sound. "We're all connected, man", sure buddy.

With AI, isn't there something similar? You use it seriously, and something suddenly clicks about what's now possible. The interface between intention and execution has changed. You feel it before you can articulate it. And then you try to explain it to someone who hasn't engaged, and you sound either manic ("I am a god now") or underwhelming ("it helped me clear my inbox").

In both cases, the experience creates knowledge that doesn't transfer through description. The skeptic can always say "it's just serotonin agonism" or "it's just token prediction" and they're correct at the mechanistic level while completely missing what happened at the experiential level.

I'm sure you had a friend or someone from school who did psychedelics for the first time at a young age, and suddenly they've found the answer. The experience is so vivid, so obviously more real than regular life, that it feels like a revelation of how things actually are. Sometimes, it metastasizes into a new framing: this is the truth, and sober life is somehow lesser, a falling away from the truth.

And if you've done enough psychedelics in your life and undergone some real proper fucking washes, and you met like a 20yo who took acid for the first or second time, in the outskirts of a rave, and you see and hear from them these exact revelations, as their mind and perception of consciousness and its nature, and life, changes in front of you, you get it too... But it's different, isn't it? But with more experience, something shifts. And it's hard to articulate because it's not like you learn new propositional content. It's more like... the frame changes.

You've already learned that the insight was never separate from life--it was always about life. The dissolving of self-boundaries, the sense that everything is connected, the vividness. They're showing you something about this, the thing you're already in.

You did the thing, you integrate its revelations, and you return to normal life. You might be more open, more aware that your model of reality is a model. But you're still living your life. The revelation doesn't do the work for you.

The AI capability shift feels similar. The access is real. The tools are finally working. But there is this fantasy of lying on a beach while it handles everything, getting a notification when everything's sorted... I guess that vision is coherent but somehow thin, existentially?

Humans have spent a long time building identities around competence, craft, and the slow accumulation of skill. If the skill becomes trivially accessible, what happens to the identity? And I think this question is a real thing people are already having trouble with.

Some people are thriving with these tools because they had taste and vision but lacked technical execution. Others are finding that the technical execution was the thing they were good at, and that's fucking them up.
I don't really know. But what I do know is this: Kind of like with psychedelics, the first real experience with AI changes your priors. But you have to figure out what to do with that. The insight is definitely real, but the integration is a separate problem. (that requires honing a skill, by the way)

So work on the integration, but try not to make it existentially thin.

[0xAyuki @0xayuki]

It really was that simple to follow 4chan anon's prophecy

[0xAyuki @0xayuki]

Using 1000 wallets to automate farming Monad testnet finally paid off 🙏

[0xAyuki @0xayuki]

@monad $MON launched on Hyperliquid, with 100B total supply, currently about $20B FDV.

[0xAyuki @0xayuki]

@alecTrading Interested!

[Kyle @zeroxkyle]

Excited to announce our lineup for @DeFianceCapital's Pitch Day!

Join us in hearing the best in the industry pitch, & a short roundtable afterwards!

This is an invite-only gathering, but we have limited spots available - DM if interested.

[Kyle @zeroxkyle]

https://t.co/dzO0s4HCDO

[0xAyuki @0xayuki]

Been using @elfa_ai to speed up my research recently.
With over 100s of new tokens daily, and millions of existing ones, information overload is something I struggle with.

Why I like Elfa's tg bot:
- has the most up to date information compared to other LLMs, even for tokens that just launched. For example, when @icmdotrun TGEd, it was easy for me to get an overview of what they do and who is behind them

- Covers blindspots on information I might have missed, with sources that I can verify

- Aggregates info from both Twitter and public telegram groups

Surprised that the tool is completely free to use, do check it out here: https://t.co/9jinTGCS2O

[0xAyuki @0xayuki]

Pretty decent results, thanks @lawmaster @TheBlock__ for the fun quiz!

[cookies (🍪,🍪) | 饼妹 @jinglingcookies]

overview of the robotics industry

[0xAyuki @0xayuki]

@0x_mailman mailman outperforming 🥳🥳🥳🥳

[0xAyuki @0xayuki]

Consensus amongst traders is "fuck trump for announcing this on a Sunday"

Higher 🚀

[0xAyuki @0xayuki]

> Have a 200M underwater short on binance
> publicly long 160M with a liquidation below ur short entry
> profit

[Hyperdash @hypurrdash]

0xe4d... is definitely the main character today

7m in collateral, 97m ETH long and 21m BTC long

40m position on GMX as well

god bless

https://t.co/ytxJLMsQs4

[0xAyuki @0xayuki]

Read my full post on substack! https://t.co/9n3Ew5GNRP

[0xAyuki @0xayuki]

Just published an in-depth game theoretic analysis of Berachain’s innovative Proof-of-Liquidity (PoL) model. Let’s dive into how it aligns incentives and secures the network through active liquidity provision.

[Delphi Digital @Delphi_Digital]

Since the inception of Delphi, research has been embedded into our DNA.

Today, we are excited to announce the Delphi Research Challenge–your chance to showcase your skills, win prizes and interview at Delphi!

Prizes and details on how to enter below.

[0xAyuki @0xayuki]

In summary, Berachain’s PoL is redefining blockchain security by merging consensus with active economic participation. This innovative approach drives a vibrant, self-sustaining DeFi ecosystem