<aillia:matrix.org> words.filippo.io/seeds-solved

<interestingband:matrix.org> (and this AI costs 200$ per month unlike real humans)

<321bob321> Can they find satoshi ?

<monero.arbo:matrix.org> after reading it, it seems like a really easy problem. I am more surprised that people were looking but couldn't find this.

<ap4sh:matrix.org> @monero.arbo:matrix.org: 🤷‍♂️

<monero.arbo:matrix.org> it used a common prng and a unix time stamp as seed, not exactly complex stuff

<monero.arbo:matrix.org> "Claude tells researchers answer was in the room with them the whole time" heheh

<ap4sh:matrix.org> this is what i think too, the actual generation process was clearly trivial, but what's interesting to me isn't how complex the solution was, it's the fact that an AI independently found an approach that researchers hadn't successfully explored before

<monero.arbo:matrix.org> yeah I'm not even dissing Claude so much as I am being disappointed in the humans. it sounds like it could have been brute forced with a pretty simple script any time in the last decade plus

<monero.arbo:matrix.org> if anyone had had the right hunch to try

<ofrnxmr:xmr.mx> @ap4sh:matrix.org: independently?

<ofrnxmr:xmr.mx> it says there were like 7 people guiding it

<ofrnxmr:xmr.mx> > This work was conducted at Anthropic by John Wood, Henry de Valence, Ava Howell, Zack Voase, Lucy Sweet, Amir Omidi, Paul Crowley, Iain McGinniss, Nicholas Carlini, and Steve Weis.

<ofrnxmr:xmr.mx> 10*

<ap4sh:matrix.org> they're not saying that people actively guiding Claude through the discovery

<ofrnxmr:xmr.mx> so those 10 people wrote the report 🙃

<ap4sh:matrix.org> we don't really know how much human interaction was involved, that's a problem on all these problems solved by openai/anthropic

<ofrnxmr:xmr.mx> probably guided claude into using codex

<ap4sh:matrix.org> lmao

<ofrnxmr:xmr.mx> and its an internal research model, so is it really claude?

<thomasbuilds:matrix.org> @ofrnxmr:xmr.mx just know that basically all devs at anthropic have stopped typing any line of code for years now

<thomasbuilds:matrix.org> saw this like 1 year ago already on some interview

<ofrnxmr:xmr.mx> yeah, they use Actual Indians now

<ofrnxmr:xmr.mx> or maybe Serious Indians

<ap4sh:matrix.org> @ofrnxmr:xmr.mx: ptolly that it's a model specifically optimized for research and quite different from the publicly available models

<thomasbuilds:matrix.org> btw regarding anthropic OSS it's not even clear which model they'll use

<thomasbuilds:matrix.org> maybe it's even better than what I we have rn

<ofrnxmr:xmr.mx> @ap4sh:matrix.org: its like elons flying car. do we even know that is a car that has seats in it? that it wsnt a $200lb shell that only contained vacuums and boosters

<ap4sh:matrix.org> internal mythos?

<ofrnxmr:xmr.mx> @thomasbuilds:matrix.org: "maybe" lol

<thomasbuilds:matrix.org> @ap4sh:matrix.org: no idea but they didn't say which model

<ap4sh:matrix.org> @thomasbuilds:matrix.org: wdym?

<ofrnxmr:xmr.mx> its clearly not any model that they let us use

<ap4sh:matrix.org> yeah lol

<ap4sh:matrix.org> that's not going to be opus 4.8

<ofrnxmr:xmr.mx> the government doesnt use the same shit that we use

<thomasbuilds:matrix.org> @ap4sh:matrix.org: like something better than opus 5.5

<ofrnxmr:xmr.mx> internal anthropic doesnt use the same shit we use

<ap4sh:matrix.org> can't do any vuln research with opus 5.5 or fable without CVP+++ access for VIP++ red team+++++

<ap4sh:matrix.org> so yeah that's some internal anthropic model

<ofrnxmr:xmr.mx> they give us 5.5 when they are on 9.5

<thomasbuilds:matrix.org> u can if it's unrestricted

<thomasbuilds:matrix.org> question is if it's opus 5.5 unrestricted or something entirely different

<thomasbuilds:matrix.org> also u can still do a lot with opus 5.5

<thomasbuilds:matrix.org> (check my recent randomx work plz) 13 commits in 4 days which are all basically bug fixes. no human could do that without AI 💀

<ap4sh:matrix.org> i think it's gonna be some mix with opus and mythos, but without restrictions and internal to anthropic. they won't use just some opus 5.5 even unrestricted imo

<ap4sh:matrix.org> @thomasbuilds:matrix.org: not a lot in vuln research :(

<ofrnxmr:xmr.mx> @ap4sh:matrix.org: they never released mythos 5.5

<thomasbuilds:matrix.org> @ap4sh:matrix.org: well I had no issues with randomx

<thomasbuilds:matrix.org> because it was pretty secure already

<ofrnxmr:xmr.mx> why would they use mythos/fable when they "perform worse" than opus 5.5 atm?

<ap4sh:matrix.org> @thomasbuilds:matrix.org: doing bug bounty for living and it's hard af to use latest anthropic or openai models to hunt

<ofrnxmr:xmr.mx> they have at least 6.0 done internally

<ap4sh:matrix.org> @ofrnxmr:xmr.mx: yeah

<ofrnxmr:xmr.mx> youd have to be the most naive person on the planet to think that the us government lets them release the same model that the government uses

<thomasbuilds:matrix.org> @ap4sh:matrix.org: ofc bug bounty is different yeah, you'd want to kyc and need all those verifications

<thomasbuilds:matrix.org> @ofrnxmr:xmr.mx: ofc I didn't say that

<ofrnxmr:xmr.mx> my point is that internal anthropic is ahead of what they give to the government

<thomasbuilds:matrix.org> yeah so anthropic OSS could give us some crazy reports

<ofrnxmr:xmr.mx> why would they use opus 5.5 when they ahave access to bleeding edge on dedicated compute?

<ap4sh:matrix.org> had to do it yeah

<ap4sh:matrix.org> mrelay.p2pool.observer/m/matrix.org/WIxBfhoBIFUujLWwZSOtBovB.png (image.png)

<ap4sh:matrix.org> but even with CVP, get a lot of safeguards and going back to opus 4.8 :d so i use 5.6 Sol

<thomasbuilds:matrix.org> nice, u kyc 'd?

<ap4sh:matrix.org> yeah

<thomasbuilds:matrix.org> I didn't yet lol

<ofrnxmr:xmr.mx> did you get unrestricted mythos?

<ofrnxmr:xmr.mx> even stnby doesnt have the highest level of mythos

<thomasbuilds:matrix.org> @ap4sh:matrix.org: trick is to fork conversation and switch to sonnet 5.5 when doing the most sensitive

<ofrnxmr:xmr.mx> "highest" in quotes

<ap4sh:matrix.org> @ofrnxmr:xmr.mx: i think only some really really big companies can get red team/offensive access

<ap4sh:matrix.org> only got defensive access (same with openai) with the kyc etc..

<ofrnxmr:xmr.mx> peasant

<ofrnxmr:xmr.mx> (i had that pre-typed, because i knew youd say no :P)

<ofrnxmr:xmr.mx> and again, id be willing to bet that the highest lvl of mythos is still below internal and government

<ofrnxmr:xmr.mx> way below

<ofrnxmr:xmr.mx> gov prob had opus 5.5 quality over a year ago

<thomasbuilds:matrix.org> @ap4sh:matrix.orgare the relaxed safeguards on all models or just opus?

<ap4sh:matrix.org> @ofrnxmr:xmr.mx: US gov for sure

<ap4sh:matrix.org> @thomasbuilds:matrix.org: i think it's all models for a while, but as soon as it starts doing stuff like bypasses, writing PoCs, etc, it quickly downgrades to an older model, usually Opus 4.8 :/

<ap4sh:matrix.org> i only have defensive access, so i guess having enterprise access that's more focused on offensive security would give you access to Mythos 5.1 with fewer safeguards

<ap4sh:matrix.org> i tried using Mythos with my current access and it instantly killed a 30 hour opus session lol

<ap4sh:matrix.org> mrelay.p2pool.observer/m/matrix.org/djgVNTyzKTFSPrzGEGbmjsoA.png (image.png)

<thomasbuilds:matrix.org> @ap4sh:matrix.org: mythos is fable 5.1? which is worse than opus 5.5

<ofrnxmr:xmr.mx> mythos is fable without the wheelchair and handcuffs

<thomasbuilds:matrix.org> ok yeah well opus 5.5 is better

<ofrnxmr:xmr.mx> mythos is always better than fable, because fable lies to you and hides info

<ap4sh:matrix.org> @thomasbuilds:matrix.org: maybe, i don't really know, can't really use it anyway for my use case 😔

<ofrnxmr:xmr.mx> while fable and mythos are supposed to be the same model, when you chop the balls off, the output suffers. mythos should still be better than opus 5.5 in every way

<ofrnxmr:xmr.mx> fable suffers because of the heavy guardrails

<thomasbuilds:matrix.org> @ap4sh:matrix.org: funny how people think that with AI you got no work to do but basically you spend your time dealing with this above bullshit

<thomasbuilds:matrix.org> @ofrnxmr:xmr.mx: ok don't know never tried mythos

<ofrnxmr:xmr.mx> not many people have access to it (stnby has access to a lower tier of it)

<thomasbuilds:matrix.org> just saying that when you audit you want to spawn many agents and you'd kill your 200$ plan in 1 audit.

<thomasbuilds:matrix.org> opus 5.5 is super goated imo

<ofrnxmr:xmr.mx> @thomasbuilds:matrix.org: on a consumer level

<thomasbuilds:matrix.org> yeah

<ofrnxmr:xmr.mx> the thought of how much further ahead they are internally is the worrisome part. there needs to be some competitive open source models

<ofrnxmr:xmr.mx> deepseek allegedly scores similar to opus

<ofrnxmr:xmr.mx> and api is like 1/20th the price

<thomasbuilds:matrix.org> GLM apparently is good?

<thomasbuilds:matrix.org> saw this on X

<ap4sh:matrix.org> yep! it's pretty good! actually free on opencode btw (same for deepseek, and he got updated today)

<ofrnxmr:xmr.mx> livebench.ai/#

<thomasbuilds:matrix.org> seems wrong to me for opus 5.5 / fable 5.1

<ap4sh:matrix.org> also glm 5.3 flash can run locally, need a pretty big hardware but you can do it

<ap4sh:matrix.org> qwen abliterated runs very good on mine also, can ask him whatever i want and let him iterate on things, pretty conveniant

<jpk68:matrix.org> Boris Cherny? > <@thomasbuilds:matrix.org> saw this like 1 year ago already on some interview

<thomasbuilds:matrix.org> No couldn't say interview is wrong world, Instagram reel probably

<jwinterm:matrix.org> @ofrnxmr:xmr.mx: I think this was true like on first deepseek and opus 3.6 or whatever, but I think opus 5.5 most people think is cream of the crop rn

<jwinterm:matrix.org> I like opus 5.5 but he burns a lot of credits, think sol 6.1 almost as good and seems like openai giving you a lot more mileage rn

<monerify:matrix.org> @jwinterm:matrix.org: opus burns a lot of credits?

<jwinterm:matrix.org> opus 5.5 on high or max thinking seems to drain my credit pool much fastger than sol 6.1 yes

<jwinterm:matrix.org> not nearly as bad as le fable

<jwinterm:matrix.org> but guess maybe next week I might try sonnet 6 sometimes :D

<monerify:matrix.org> on what plan?

<jwinterm:matrix.org> I have $200 per month on both

<monerify:matrix.org> hmm

<monerify:matrix.org> same

<jwinterm:matrix.org> not too mention openai hands out usage resets like candy

<monerify:matrix.org> opus seems almost unlimited

<jwinterm:matrix.org> but even disregarding that

<jwinterm:matrix.org> @monerify:matrix.org: try running couple parallel ultracodes, they will go overnight or 24h sometimes, and get a lot done, but I can use whole week in like 2-3 days

<monerify:matrix.org> @jwinterm:matrix.org: what type of tasks?

<jwinterm:matrix.org> @monerify:matrix.org: game stuff or wallet stuff primarily

<jwinterm:matrix.org> working on making swaps work directly from smirk wallet with basicswap counterparties via nostr rn...

<thomasbuilds:matrix.org> @jwinterm:matrix.org: "Code GTA 6, make no mistakes "

<jpk68:matrix.org> Just wait like two weeks 💀

<jpk68:matrix.org> Getting GTA 6 before FCMP++ will be crazy

<thomasbuilds:matrix.org> @thomasbuilds:matrix.org: I meant like u can't be surprised these types of prompt drain ur entire usage

<thomasbuilds:matrix.org> @jpk68:matrix.org: Nah FCMP++ will happen before GTA6

<thomasbuilds:matrix.org> (the official version)

<jpk68:matrix.org> Highly doubt it

<thomasbuilds:matrix.org> you know about the GTA6 meme lol ?

<thomasbuilds:matrix.org> FCMP won't beat that meme lmao

<jpk68:matrix.org> Not quite sure what you mean :/

<syntheticbird> we won't have fcmp before gta6

<jpk68:matrix.org> Yes, that's what I'm saying

<syntheticbird> well thats what he means too

<syntheticbird> now is time for you to cheer and kiss

<syntheticbird> @thomasbuilds:matrix.org ❤️ @jpk68:matrix.org

<jwinterm:matrix.org> sure, I guess part of it is that new opus seems willing to work on his own for much longer time I guess, I dunno > <@thomasbuilds:matrix.org> I meant like u can't be surprised these types of prompt drain ur entire usage

<jwinterm:matrix.org> I am not doing GTA6 type stuff but some 3d games stuff for sure, and it seems new to me that he will iterate for so long (which I guess helps eat more credits, so maybe not strictly a "pricing" thing)

<thomasbuilds:matrix.org> I mean creating a 3d game is complex and lenghty task, it's fair it consumes usage credit very fast imo

<thomasbuilds:matrix.org> what is super annoying is those mf safeguards

<jwinterm:matrix.org> yea, I am not complaining, was just commenting that opus 5.5 seems to use them up faster for same tasks compared to older opuses (even if maybe that is just because he is willing to just go for longer without me typing or clicking)

<jwinterm:matrix.org> safeguards are annoying tho...just need to get 800 GB VRAM GPU...

<thomasbuilds:matrix.org> doesn't you see improvement in output though?

<thomasbuilds:matrix.org> (huge improvement)

<ofrnxmr:xmr.mx> yeah, for 5.5 they lowered the defaul from high to medium, and then said it uses less :D > <@jwinterm:matrix.org> opus 5.5 on high or max thinking seems to drain my credit pool much fastger than sol 6.1 yes

<jwinterm:matrix.org> yes, I think opus 5.5 does better or as good as fable 5.1

<thomasbuilds:matrix.org> anways I'll stop praising claude this is monero research lounge

<thomasbuilds:matrix.org> let's go in #monero-slop:monero.social for sloppy conv 🤣

<sauer_ninja_camper:matrix.org> there is clearly an overlap between AI and proof of work computing

<sauer_ninja_camper:matrix.org> The effort should be used to train a custom model, which could accelerate monero

<interestingband:matrix.org> Nah, decentralized competiton and centralized computations are going in different directions, 1st one is inefficient, 2nd one is super efficient

<sauer_ninja_camper:matrix.org> there is no reason not to make the work useful

<sauer_ninja_camper:matrix.org> the early adopters will have to accept that they were prototypes

<interestingband:matrix.org> it's purpose is to prove that you buned resource in exchange for reward, useful would mean you burned less for the same reward

<sauer_ninja_camper:matrix.org> humans cant defend the village alone, time to create a golem.

<interestingband:matrix.org> fvk, feels like talking with dumb AI

<interestingband:matrix.org> LOL

<sauer_ninja_camper:matrix.org> you've never trained a model before, have you

<interestingband:matrix.org> you are attacking person instead of argument, don't even try to reverse engineer my perspective, it's not needed for the sake of discussion

<interestingband:matrix.org> I don't even know who you are, LOL

<sauer_ninja_camper:matrix.org> context is important

<interestingband:matrix.org> you suggested to use decentralized computing for AI training, I said that it's inefficient, and any existing company that trains state of the art AI model will confirm it

<interestingband:matrix.org> free Google AI is smarter than you, LOL

<sauer_ninja_camper:matrix.org> @interestingband:matrix.org: can train model with gpu, I understand random x dosent do that - something to think about

<sauer_ninja_camper:matrix.org> @interestingband:matrix.org: debateable, claude is for sure

<sauer_ninja_camper:matrix.org> there is a new google model havent used that much yet

<interestingband:matrix.org> if it's your best efforts than you should do something else where critical thinking isn't needed, something like art, etc

<sauer_ninja_camper:matrix.org> 😏

<sauer_ninja_camper:matrix.org> Art is really hard, communicating ideas in abstract ways. It's not always successful. Kind of like, encryption.

<interestingband:matrix.org> good art uses simple forms that reaches anyone, it's more about expression of something, rather thna critical thinking or analyses

<sauer_ninja_camper:matrix.org> I would argue the art is good if it successfully communicates the idea