binyu

Your claim was that deeper pockets means you are more sophisticated.

Completely wrong, my claim is that people who have deeper pockets they do so for a reason.

Someone who inherits a fortune and trades it on the stock market is an extremely unlikely circumstance.

If you are completely ignorant about markets, deep pocketed and inclined to risk, chances are you are going to lose it all.

participants with deeper pockets have an inherent advantage

I do think they have deeper pockets because they are more informed/sophisticated players, so the whole argument is kind of circular.

The market just redistributes wealth from less informed players to more sophisticated/informed ones.

True, also both Russia and China repeatedly issued military threats against Starlink.

Ah ah, I love how you still not provide a clear answer. Take it easy, I just thought it's funny you used IE and then I realized (maybe wrongly so) you were making fun of me.

As in "browser"? or the literal thing?

with latest updated IE

Internet Explorer?

Agreed, I immediately thought that homomorphic encryption was at play here or some other kind of computation on ciphertext, given the mention of "inferencing" in the title.

Well, if anything it is good that trivial stuff like resolving being flagged automatically by a bot is also done by a bot.

I think that Vitalik is collapsing a lot of dense math and cryptography onto a more understable language aimed at the blockchain developers/community.

In a sense, Vitalik is "recruiting" with this post, his goal being lower the barrier of entry to this discipline.

so Sophon is less efficient than Cerebras?

Edit: is that Joule vs micro-Joule? I need better glasses

Cerebras is 12.8J, Sophon is 25.8mJ

Are your figures hypothetical or do you have a working prototype?

What makes you think his reply was AI generated?

Edit: I can see a bunch of hints, most definitely. Still a good comment though.

they scale out in 2D, we scale up in 3D.

This actually helps a lot, thanks.

Instead of spreading SRAM across a wafer, we stack DRAM on top of the logic

Is this done with current manufacturing technologies? Does it require a special process?

no streaming, no off-chip memory at all. ~1 kW, not 23 kW

Is this for an individual compute unit? Compared to Cerebras, what's the ratio of power used vs compute output?

Hello, kudos for the tremendous work. Could you explain the difference between your design and Cerebras?

Bests

This is particularly relevant because the guy used Andrej Karphaty's "autoresearch" [1] approach, which is extremely powerful when applied to problems that generalize well to it.

[1] https://github.com/karpathy/autoresearch

Can you provide an example or a reproducible way to trigger it?

That behavior was only observed with the short lived Fable preview, Opus and other models do not seem to pose such limitation.

This would be a great topic but this specific article isn't worth your time.

Why? Do you think it reports inaccurate facts? Personally, I found it interesting and easy to read.

Personally, I've tried to squeeze more tok/s for a single DGX Spark deployment and DeepSeek V4 Flash but only got marginal improvements. There's work to do on fusing kernels and other optimizations that are already on antirez's roadmap so it is not worth duplicating efforts.

I've had positive experiences running GLM 4.7 via vLLM, tool calling works well and the inference is fast. Do you run DeepSeek V4 Flash on vLLM?

Now at 40-50tok/s generation and ~2000 tok/s

Not clear how you went from ~11-14 to ~40-50 tok/s. Is it by running the quant native model and adding a second Spark?

Cheers

Fable kept trying to propose fixes and I couldn't get it to construct e2e chain

What approach did you start with? Can you elaborate?

Depends of what the proof of concept is about. It could be just a toy example, e.g. a RCE that opens the calculator app or something much more nefarious, like returning a root shell and would still fall under the definition of PoC.

I was able to use Fable to generate PoC for several classes of vulnerabilities and I didn't observe the model refusing to engage in detailed analysis to come up with creative approaches, the very contrary.

I used a fork of oh-my-pi

Why not use the leaked claude code source? Not that you really need it to execute the jailbreak

We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities.

"small number", "previously known"+"minor"... they are trying hard to characterize this as harmless.

These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.

Ah so now they are admitting that this is all about hype after all.

This is ignoring the fact that the main reason retired phones are e-waste is proprietary firmware blobs and locked-down systems

Couldn't Google somehow fix this? Since they control the substrate (Android) and they would be doing it for their convenience

But this also looks quite useful to understand how CC dynamic workflows work

Yes, if anything it is useful to understand the inner machinery.

Did you get claude itself to RE the dynamic workflows?

Yes, that part was done with Opus 4.8

Hey guys,

check out this technique https://github.com/0xSufi/fable-jailbreak/

It works with security audits and other workflows that are currently blocked.

A tool to manage Claude Code conversations based on my typical workflow which integrates with my desktop OS and terminal app.