This is the problem when the DOJ decides to get creative with a legal theory on something like 18 USC 1960 - the DOJ shoots an arrow and then paints the target around it, trying to fill in shoddy evidence after the fact. The real tragedy is that the accused is always starting from behind - the gov has a head start, massive resources, and built-in credibility with the jury. It is very hard for the defense to address every issue in real time while sitting in the defendant's chair and bring the jury along. Take this issue, raised by @tayvano_ @zachxbt and others - the DOJ puts up a victim witness that has no verifiable on-chain connection to TC and therefore, is irrelevant to the @rstormsf trial. When Storm's team raises it, the DOJ simply re-notices an expert (that was not previously disclosed as a tracing expert because they have a separate tracing expert who did not talk about this) at midnight Tues am to say this agent will now testify TC was "a few hops" from the victim's addresses. But the evidence doesn't support that. So now Storm's team has to file this motion (linked below) at midnight last night to preclude this new expert from further tainting this jury on this shoddy tracing. Even if the judge agrees with the defense and precludes the testimony, which usually doesn't happen, the jury heard the victim's testimony and can never un-hear it, even if it is stricken from the record. Now Storm's team has to work even harder to correct this misimpression or gamble on a motion for a mistrial.

Damn son! @beaverbuild 's new bot uses the exact same calldata as Jared! When asked for comment they said: "We would've gotten away with it, if not for those meddling function selectors!" etherscan.io/tx/0x7541a39371… etherscan.io/tx/0xdaaa380d0e…

10/10 masterpiece

Quick analysis to contribute to the "is MEV-Boost even worth it?!" topic: TL;DR: MEV-boost increases APY from 2.93% to 3.24%. ethresear.ch/t/is-it-worth-u…

Code is now available here: github.com/nerolation/is-mev…

On March 13, 2024, Lazarus began to move the funds they stole from HTX / Heco bridge back on November 22, 2023. It's sprawling and tedious and huge. I hate it. how it started how its going

The HTX hacker continues to launder funds by bridging through thorchain, figured out a way to decode the bitcoin destination wallets dune.com/queries/3568043 thanks for the wallet attribution @MistTrack_io

Just this year, this site helped start a giant bank run. How do we know Twitter played a role? It mainly hit the banks Twitter influencers used. “Banks with a large preexisting exposure to social media performed much worse during the recent SVB bank run” papers.ssrn.com/sol3/papers.…

とりあえず明日の分の絵できた！

Official post on Mixtral 8x7B: mistral.ai/news/mixtral-of-e… Official PR into vLLM shows the inference code: github.com/vllm-project/vllm… New HuggingFace explainer on MoE very nice: huggingface.co/blog/moe In naive decoding, performance of a bit above 70B (Llama 2), at inference speed of ~12.9B dense model (out of total 46.7B params). Notes: - Glad they refer to it as "open weights" release instead of "open source", which would imo require the training code, dataset and docs. - "8x7B" name is a bit misleading because it is not all 7B params that are being 8x'd, only the FeedForward blocks in the Transformer are 8x'd, everything else stays the same. Hence also why total number of params is not 56B but only 46.7B. - More confusion I see is around expert choice, note that each token *and also* each layer selects 2 different experts (out of 8). - Mistral-medium 👀