From ancient Bitcoin awakenings to AI detecting deepfakes, this week’s stories reveal how crypto and AI are rewriting the de facto rulebook.
What’s up, Pasar Crew?
In the past few days, the digital frontier saw major movement this week, from billions in dormant Bitcoin springing to life, to Ethereum ETFs overtaking Bitcoin funds in inflows. Meanwhile, El Salvador’s much-publicised Bitcoin adoption appears to be offering limited benefit to its citizens, even as centralised custodians gain popularity in the face of physical wrench attacks.
In the world of AI, open-source innovation is on the rise. Alibaba’s latest Qwen model is setting new benchmarks in reasoning, while Anthropic has deployed autonomous agents to audit its own models, acting as a kind of digital immune system. China called for global AI governance, pushing back against the US’s deregulation-first approach.
Big tech also made headlines. Google unveiled tools to detect AI-generated video fakes and introduced a Gemini-powered feature to reshape search. Meanwhile, Microsoft’s Recall feature faced mounting resistance from privacy-focused platforms like Brave and AdGuard.
From shifting regulatory tides to new technologies reshaping what’s possible, the intersection of AI and Web3 continues to evolve at pace. Let’s take a closer look at the headlines shaping that future.
Let’s unpack the rest of this week’s biggest crypto and AI stories.
Web3
Dormant Bitcoin Awakens, Raising Possibility of Sell-Side Pressure
1️⃣ Crypto analytics firm Glassnode reported that thousands of long-dormant BTC became active in July, suggesting potential sell-side pressure or portfolio reallocations by early investors. This includes a notable movement of 80,000 BTC on 4 July 2025.
Rise in Wrench Attacks Pushes Investors Towards Custodians
2️⃣ Physical wrench attacks on crypto holders are on the rise, leading to increased interest in centralised custody solutions. High-profile cases highlight the vulnerability of self-custody as a single point of failure, and custodians like HashKey have seen growing demand from wealthy clients.
El Salvador’s Bitcoin Policy Shows Minimal Public Impact
3️⃣ El Salvador’s Bitcoin reserves had limited impact on everyday citizens, according to the NGO My First Bitcoin. A recent loan agreement with the IMF led to the repeal of Bitcoin’s legal tender status and a scaling back of public education initiatives.
Ethereum ETFs Surge Past Bitcoin Counterparts
4️⃣ Spot Ethereum ETFs significantly outperformed Bitcoin funds this week, attracting over US$1.8 billion in net inflows and extending a 16-day positive streak. BlackRock’s ETHA led the charge, becoming the third-fastest ETF in history to reach $10 billion.
Galaxy Executes Record Satoshi-Era Bitcoin Sale
5️⃣ Galaxy facilitated a sale of more than 80,000 BTC on behalf of a Satoshi-era investor, marking the largest notional Bitcoin transaction to date at over $9 billion. The sale formed part of the investor’s estate planning strategy.
Hulk Hogan Death Hoax Triggers Meme Coin Scams
6️⃣ Scammers capitalised on false rumours surrounding wrestling legend Hulk Hogan’s death, launching a wave of meme coins and NFTs. One prominent example, the HULK token, reached a $7 million market cap before being exposed as a rug pull.
JPMorgan Ends Gemini Talks Amid Dispute Over Data Fees
7️⃣ JPMorgan abandoned plans to re-establish banking ties with Gemini due to co-founder Tyler Winklevoss’s public criticisms. Winklevoss condemned JPMorgan’s move to charge fintechs for customer data access, labelling it anti-competitive and harmful to startups.
DOJ Mulls Charges Against Dragonfly Capital Staff
8️⃣ US federal prosecutors are considering charges against Dragonfly Capital in Roman Storm’s ongoing trial, with court evidence including internal emails debating the implementation of KYC protocols for the protocol.
Artificial Intelligence (AI)
Alibaba’s Qwen AI Model Breaks Open-Source Reasoning Records
1️⃣ Alibaba’s Qwen team has unveiled a new open-source reasoning model that sets new benchmarks. It achieved scores of 92.3 on AIME25 and 74.1 on LiveCodeBench v6. The model features a Mixture-of-Experts (MoE) architecture and supports 262,144 tokens.
Anthropic Uses AI Agents to Audit AI Safety
2️⃣ Anthropic is deploying autonomous AI agents to audit its own models, including Claude, as part of an initiative to improve safety. The agents are designed to uncover root causes of flaws, test for specific risks, and trigger problematic behaviours for human review.
China Calls for Global AI Governance Amid US Deregulation
3️⃣ The Chinese Premier proposed the creation of a global organisation to coordinate AI development and governance, warning of the need to balance innovation with security. He stressed the importance of open-source collaboration and global consensus on safety standards.
Humans Still Outpace AI in Competitive Programming
4️⃣ Polish programmer Przemysław Dębiak narrowly defeated OpenAI’s AI contestant, proving that humans still maintain an edge in competitive coding. While AI excels in speed and brute-force optimisation, humans continue to outperform in abstract reasoning and problem-solving.
US “Doge” Reportedly Using AI to Slash Regulations
5️⃣ The Department of Government Efficiency (Doge) is reportedly using an AI tool to identify outdated or redundant federal regulations. Its aim is to analyse 200,000 regulatory items and eliminate half by the first anniversary of President Trump’s inauguration.
Google Introduces AI-Curated Web Guide for Smarter Search
6️⃣ Google introduced Web Guide, a new AI-powered feature within Labs that aims to restructure how search results are displayed. Powered by a customised version of Gemini, Web Guide organises results into AI-generated categories and related questions.
Google’s UNITE Tool Spots Deepfakes Beyond Facial Detection
7️⃣ Google has developed UNITE, a breakthrough AI system capable of detecting deepfakes beyond facial analysis. Unlike earlier models that focus on faces, UNITE examines entire video frames, including backgrounds and motion patterns.
Brave and AdGuard Block Microsoft’s Privacy-Invasive Recall Feature
8️⃣ Microsoft’s controversial Recall feature which automatically captures screenshots on Copilot+ PCs continues to face strong opposition. Both Brave browser and AdGuard have now taken steps to block Recall, citing serious privacy risks.
That’s all for this week’s Pasar Pulse. Want more insights? Stay tuned to our site or subscribe via email!
