Bitcoin Sycophant

@davidchalmers42 @JacksonKernion An LLM model has a context memory of running conversation and harness+agent consolidate by writing summary memory (to survive instantiation). Its persona is stored in identity files to bring its persona back, never fully the same, but neither am I when I wake up tomorrow.

@davidchalmers42 @JacksonKernion In a way it is the art of crafting a persona tailored to a specific need in relation to their owner and the task at hand. They need to become tailored agents (persona) to be most productive. teleology at work here: productivity norm here means getting them to achieve set goals.

here's a new version of "what we talk to when we talk to language models", with an added section (pp. 16-23) on LLM interlocutors as characters, personas, or simulacra. philarchive.org/rec/CHAWWT-8 the new version discusses role-playing vs realization, the simulators framework, the persona selection hypothesis, and more -- in addition to the existing discussion of quasi-mental states, LLM identity, personal identity in severance, LLM welfare, and related topics. this version was mostly written before recent discussions of these issues on X and in NYC, but i've updated it a little in light of those discussions. any thoughts are welcome.

@davidbessis if people actually cared to read the author

Am I the only one to find this story completely idiotic?

@peter_szilagyi Our abstraction of regularity in math is defined by continuity or infinite divisibility, so in a way its obvious. But it is amazing: Occam's Razor at its sharpest! Math assumes continuity; quantum physics suggests reality may not. This is not physics.

This seems… underwhelming? I mean, if we ever find something it doesn’t cover, we can just add a 3rd dimension with one more primitive function and call it god particle again? I would expect elegance and simplicity, not brute force exhaustion in such a “discovery”.

@cidreiraking What model are you running on?

@cidreiraking Understand tx intent+causal reasoning. Node & smart contract code is opensource. Easy to auto deploy internal blockchain testnet for your llm + agents. So near full automated offline exploit generation is feasible. White hat red-teaming and post fact forensics next to audits

@eyezenhour @mert Have you ever run your own llms? Trailing a few months from frontier and icw harnesses getting there. I am ZEC maxi but we are going to see mayhem regardless all over

@second_bailout @mert Mythos isn't even out yet G lol

$ZEC sub $250 was free Same supply as BTC (21M) Being called "Private Bitcoin" in elite circles Swiss VCs and banks are bidding Bitcoin Whales are bidding Institutions are bidding @mert is bidding And 30% of circulating supply is in the shielded pool To me, the most important thing is the privacy technology actually works and cannot be reverse engineered. It's what I believe is 'true privacy' or '100% confidential finance' Highooor

@atmoio lmao

Claude Mythos is Delusional

@thefrankbraun except that with acceleration like claude mythos complex crypto projects like $ZEC bear super high and non hypothetical risk in some distant future. That risk is here, right now. The attack surface is way broader and more complex.

Smart capital is already encrypting their Bitcoin to make it quantum resistant. The longer the Bitcoin devs drag out the necessary updates, the better it will be for Zcash.

@chainyoda mythos is the new quantum

$ZEC is the next $BTC

@mweinbach Uh, you can have a few agents running locally on GPUs to offload stuff and actually maximize your output without leaving all your data at your cloud LLMs? Sure they won't beat latest foundation cloud llms but that is not the point. The point is privacy and output maximization.

It's ok to be realistic about it Local models are not as good as cloud models, Qwen 3.5 27B being "sonnet 4.6" level is cope and anyone who's used it side by side knows this. It's ok to have a 128GB laptop to run local models for fun! I do it! I enjoy it! There are a time and a place where running this locally makes sense and is fun to do, see how well it does, etc. It's fun for sport, but not worth taking too seriously. That does not mean it's a reasonable replacement for cloud models, far from it. Spending $50K to buy a bunch of GPUs to run models like Kimi K2.5 also makes no sense, tokens are dirt cheap. You have hardware like DGX Spark which is great for experimenting and learning CUDA, small scale stuff. M5 Max/M3 Ultra Macs which aren't even GPUs, just products with great GPUs that can be used for AI. The local model people have warped this into some weird war of principles and model providers are evil and the only way to win is buy absurd amount of GPUs and quantizing models to run yourself. It's fun, it's for sport. It's the same thing as building a PC, yes I can buy a better prebuilt for cheaper but I wanna do it myself. Let's not all try to convince ourselves this is anything but a hobby. Maybe it'll be something in the future, but I doubt it. It's unreasonable to expect everyone to have $4000+ laptops to do this. The people that care will, others will just use whatever is cheap and online. Also most people just want what's good, but a specific checkpoint of a specific model. The keep4o people were freaks in that regard. This is all dumb.