
阿空(🐂, 🐂) 互关学习🫡
13.6K posts

阿空(🐂, 🐂) 互关学习🫡
@ResearchKONG
🐦 Web3丨美股投研 | 数据挖掘 分享早期机会 | 实盘操作 | 亏损总结


Everyone knows how to manage money online 😂




I wrote about why blockchain's killer app might not be an app at all. For 17 years we've asked people to adapt to the machine. But every computing era ends with the machine adapting to the person. AI is how that finally happens on-chain. Full read: evernorth.xyz/blog-post-06-1… This content is for informational purposes only and does not constitute investment advice. Digital assets involve risk, including potential loss of principal. Learn more about Evernorth: evernorth.xyz/blog-post-03-1…



if ur trading crypto transition to trading stocks, xyz markets on hyperliquid have low fees, i rather trade gold and commodities or indexes for 5 times less fees than any manipulated altcoin










Today NCUA joined other federal agencies to announce a proposed rule for permitted payment stablecoin issuers to establish and maintain an effective customer identification program under the GENIUS Act. Interested in learning more? Visit bit.ly/3QnzunB for details.









@federalreserve requests comment on proposal to require certain payment stablecoin issuers to maintain an effective customer identification program: federalreserve.gov/newsevents/pre…



$AMD $INTC AI Inference & Agentic AI Tailwind. Is Market finally pricing in the CPU necessity? -AMD CEO Lisa Su disclosed at the March 2026 Morgan Stanley TMT Conference that server CPU demand has "far exceeded" prior forecasts. -Meta and AMD announced a five year, $60 billion agreement anchored by 6th Gen EPYC "Venice" CPUs paired with next-gen GPUs, with workload-specific tuning for inference-heavy environments. -Intel and SambaNova are building a disaggregated inference architecture specifically for agentic AI. Intel Xeon 6 CPUs handle agentic tool execution and system orchestration. SambaNova RDUs handle LLM decode. GPUs handle prefill. -In agentic pipelines, the orchestration layer accounts for 50 to 90% of total system latency. That makes CPU performance the rate-limiting variable. Not VRAM. Not FLOPS. -Deloitte forecasts that inference will make up two thirds of all AI compute by 2026, far exceeding initial training costs.






Bought $NVDA 207.5p for .50 0DTE might go tits up















