Testimony regarding the fact that some of the stolen funds from IRS Special Agent Stephen George’s Hanfen Lin had been placed in tornado cash seemed reliable enough to prevent the defense from filing a claim of tort.

On Monday, the defense of the tornado cash trial came to mind to file a motion for the Miss Trial after saying that FBI special agent Joel DeCapua had not tracked the tracking of funds stolen from the case’s first witness to Tornado Cash Mixing Services.

However, after today’s testimony from IRS special agent Stephen George, the second witness, the second witness who prosecutors sought training to track assets on public blockchains, seems slim at best to file a complaint of dispute.

Special Agent George hired the accounting principals from LIFO (last, first out) that included Etherscan.io (Ethereum’s Block Explorer), Chainalysis’ Reactor, and TRM Labs, and spoke about how they used the TRM Lab to track stolen funds.

According to data from special agent George, approximately 149,000 USDT was sent from Lynn’s crypto.com account through a series of wallets. About a third of this amount was exchanged for 9.78 ETH (worth around $47,000 at the time) before moving the Tornado Cash on November 15th, 2021.

Special Agent George admitted that, according to initial disclosures from prosecutors, he was not tasked with tracking these stolen funds from crypto.com to tornado cash, but began to do so at the end of June or early July this year.

This defense was thorough in cross-examination of special agent George, but it seemed to struggle to find holes in his story and the notable flaws in the analytical process.

Special Agent George admitted that his analysis may not have been perfect, but said that the results of his analysis using chain analysis and tracing tools from TRM Labs independently produced similar results.

That being said, the defense shared that he plans to complete Cross Agent Special Agent George at the beginning of tomorrow’s trial.

So, there is still a possibility that he could find a way to trust some of the data he presented, but based on today’s cross-examination, it doesn’t seem possible.

See this thread in X for a more detailed explanation of what happened in court today.

According to the company’s announcement on May 7, “Mistral Medium 3 delivers frontier performance while being an order of magnitude less expensive”.

Medium 3 joins the company’s lineup of previous releases, including open-source models like Mistral 7B, Mixtral, Codestral, and Pixtral. However, unlike those earlier models, Medium 3 is not open-source, and users cannot modify or run it on their own devices.

Did you know?

Want to get smarter & wealthier with crypto?

Subscribe – We publish new crypto explainer videos every week!

What is a Crypto Wallet? (Explained With Animation)

Medium 3 costs $0.40 per million input tokens and $2 per million output tokens. Reviews by Artificial Analysis rated it as one of the best non-reasoning models available. It competes closely with options like Llama 4 Maverick, Gemini 2.0 Flash, and Claude 3.7 Sonnet.

Sophia Yang, a representative from the company, stated:

Mistral Medium 3 shines in the coding domain and delivers much better performance, across the board, than some of its much larger competitors.

Additionally, Medium 3 supports a 128,000-token context window and can process text and images in over 40 languages. This makes it useful for a wide range of tasks, including handling documents and visual content.

Currently, the model is aimed at business use rather than everyday users. It is available through LeChat, Mistral’s chatbot interface, and is designed to work with business systems. Companies can fine-tune it, connect it with internal data, and keep improving it with continuous updates.

Meanwhile, on April 28, OpenAI introduced new features to its AI chatbot, ChatGPT. What are they? Read the full story.