crypt0grapher (@7blockchains) header
crypt0grapher (@7blockchains) profile picture

Homo Ludens ♦️ Forging Quant ⚡ Chainwright thunderswift trade automata 🔮 Dispensing grimoires at Zero Lag Club 📜

EU or coffee shops worldwide
Joined September 2021
Photo by crypt0grapher (@7blockchains)Photo by crypt0grapher (@7blockchains)Photo by crypt0grapher (@7blockchains)Photo by crypt0grapher (@7blockchains)Photo by crypt0grapher (@7blockchains)Photo by crypt0grapher (@7blockchains)
Replying to @jessexai_
Thin ice zone. Government welfare and new markets don't solve the contradiction. LLMs are constant capital. Only human labour creates surplus value. No surplus value → the rate of profit goes to zero → huge questions about capitalism's survival. What society is being built?
105
Replying to @zerohedge
wouldn't the very concept of frontier be moved by that?
145
Replying to @predict_addict
Best of luck brother. Let the Phystech spirit help you in this decent endeavor. No doubts you’ll make it.
1
1
153
Replying to @ivanfioravanti
It is neither unless they can cluster as DGX
32
Wow thanks brother, missed that somehow
1
19
wdym? are they moving out of SG?
1
31
Replying to @lxe
Haha, that’s a really good one lol :-)
9
Replying to @predict_addict
Only one synapse? :) "Mathematics should be studied if only because it brings order to the mind." I bet you know who said that.
42
Replying to @ChShersh
Why? I bet it's Cursor-only now, which is VS Code on steroids. And yes, AI-only most likely; why bother? A huge fan of @jetbrains products here. Damn, I remember the first version of ReSharper... How old am I
38
Models used in the session: - grok 4.7 xhigh - claude opus 5.5 xhigh - codex gpt 6.1 sol xhigh - kimi k3 max @xai is killing it
37
Claude Opus 5.5 team lead thinks @grok is the best. The project is a multi-faceted, low-latency blockchain software with multiple safety layers. It's a shame @OpenAI Codex keeps f-ing up code reviews (no smart contracts, no private keys, no reverse-engineering - one of us doesn't define well what cybersecurity is). "Grok: best value for bounded work". Can't agree more! The opinions in the screenshot are not my own. 🤖
1
59
Educated people are making up stories about AGI conquering the world, but we're not even close to a thinking machine. It's literally comparing matrix multiplication of predefined weights with the ms changes across hundreds of trillions of connections to create a single thought.
10
Replying to @predict_addict
Someone, finally. Still, it's not convincing and never will be. Marketing consumption is way easier than learning the basic bits of math and biology.
1
124
Replying to @systematicls
True, weekly limits are gone in a day, that’s hardly usable. LLMs are getting more and more expensive, more and more cabals though! Still I think the new tower is a better deal than the one from Anthropic that tried the 17% cut to look like a 25% benefit.
1
47
Replying to @digitalix
Great job with the tests! This was not an easy video to shoot, lots of hard work went into. It's a pity that 60+ tks/s is somewhat of an entry ticket to real usage (e.g., coding). The trick is you CAN link 4/8/16 DGXs and make it a real competitive tool, but the Mac stays on the desk as-is (still being an amazing machine though).
394
Replying to @predict_addict
100%. All Dr Chan's books are great: simple ideas, clearly explained - an amazing recommendation for beginners and non-beginners
1
811
got the same on my second claude account when i tried to upgrade it, they really just ignore appeals. I filed one, but “after careful review and considerarion” they refused. It’s so absurd, given my first account is of the same usage and being heavily loaded without a “careful review” so far. something is broken at @AnthropicAI
1
73
looking forward to the grok 4.8 really, 4.7 is the best roi obviously, but it’d be awesome to make it a bit more autonomous
67