🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Looks like Google got its “AI-mojo” back from responses from Google-IO

Spoke at Zuitserland on May 12 about hybrid cyber+bio+AI risks—building on earlier work with Anders Sandberg . Just days later, pleasantly surprised to see Anthropic activated ASL-3 citing similar concerns. LLMs amplify dual-use risk. anthropic.com/news/activatin… #AI #biosecurity #zk

The more i think about hybrid risks, more i see evidence pointing in the direction sean.heelan.io/2025/05/22/how… < O3 used for finding zero day …

At Zuitzerland , I warned LLMs would soon discover zero-days. Now it’s real: CVE-2025-37899 found via O₃, an LLM-based tool. The AI–cyber hybrid risk isn’t theoretical anymore. It’s live. We need verifiable compute + crypto-secure infra. Now. sean.heelan.io/2025/05/22/how…

It’s amazing how Anish Mohammed warned about this earlier this month at Zuitzerland! 👏👏

If you are in AI & is interested in x-risks, don't miss this one! AI has already levelled up to a serious threat level! @anthropic itself have marked Opus 4 under ASL 3! Sit up & watch! 😲

Come join us for the d/acc day and hear from vitalik.eth and many others d/acc-ing. Join us on the 11th.

Claude, Grok, Gemini, GPT — the Four Horsemen of the LLM wave. They’ll rewrite your docs, automate your life, play therapist, and even happily pick a fight over breakfast.

Sat in corner of coffee shop working off my laptop with apple vision, getting strange stares - most dont recognise that you could see them ....

macOS 26 sherlocks Docker with support for Linux containers appleinsider.com/articles/25/06… < native support for containers good to see, not sure if it adds anything more to the possibility of 512 GB of memory- making mac favourite for running LLM’s …

Looking forward to giving a talk at d/acc day

Revisiting the talk at Zuitzerland at d/acc day in Berlin

#ML paper: GUARD—guided unlearning for LLMs! Uses data attribution to selectively forget, preserving retained knowledge. No more unnecessary utility loss! outperforms baselines on TOFU benchmark. Read more: arxiv.org/abs/2506.10946 #AI #Unlearning

Interesting architecture for multiagent systems , and a good general read on topic anthropic.com/engineering/bu…

Warren buffet said “You don’t have to do exceptional things to get exceptional results.” - having been in various tech ecosystem mostly from start, think Web3 might be the closest case that applies

Everyone is interested in “decentralised “ machine learning,, here is an instance of “asynchronous decentralised machine learning “ perplexity.ai/page/chinese-a…

Some talks are good, some are great, some are exceptional, this one from Andrej Karpathy - is definitely one of them - his talk at startup school - software in the era of ai youtu.be/LCEmiRjPEtQ?si…

Ran into “prof of computer science “ in local university- using claude code to try find her cheating students, while at coffee shop - both cheating and detection using LLM leverage 😂😂😂😅

AI is breaking the internet’s memory via Dataconomy dataconomy.com/2025/06/18/ai-… < wondering if this is “rent seeking “ and wondering if there are mechanisms so that we have a “win-win”

If the numbers are to believed #crypto adoption is higher than 2024 and #stablecoin volume is also up.... h/t a16z crypto