Latest posts from @Kay2289123 on X (Twitter)
来自硅谷的投研团队
成员来自斯坦福、沃顿商学院
懂技术的投资者，懂投资的工程师
致力于让模型更聪明、更快、更便宜
What they post about
The account focuses on insights related to AI training and inference, highlighting developments in the tech industry, particularly in Silicon Valley. It discusses investment strategies within the tech sector, analyzing market trends and stock performance of companies like Marvell and Micron. The account also delves into the significance of token consumption in AI and the evolving landscape of AI hardware. Overall, it offers a professional perspective on navigating market dynamics and investment opportunities.
Recent posts
-
tweet
AI Inference Explainer Examines Speed and Capacity Trade-Offs
A technical post explains how AI inference systems balance response speed, throughput and GPU use. It outlines metrics such as time to first token, describes how requests are scheduled and batched, and argues that growing demand could create opportunities for computing-infrastructure companies.
View on X (Twitter) → -
tweet
Investor Claims AI Capacity Is Already Spoken For
The author says they deleted an earlier Cerebras post after a friend advised against sharing private information. They claim Jane Street and others have already secured available Cerebras and OpenAI capacity, and say they bought shares during last Friday’s dip ahead of the next trading session.
View on X (Twitter) → -
tweet
OpenAI Usage Limits Draw Criticism After DevDay
The author says OpenAI’s DevDay fell short of expectations and criticizes reduced usage limits. Citing their experience with paid plans, they argue that computing and networking constraints are limiting service capacity and speed, reinforcing their long-term interest in companies supplying that infrastructure.
View on X (Twitter) → -
tweet
Desktop AI Tool Could Streamline Investment Research
The author sees two benefits for investment research: an AI could work across reports, local PDFs, websites and data sources on their computer, while offering a potentially less costly way to automate desktop tasks than OpenAI’s Computer Use.
View on X (Twitter) → -
tweet
Open-Source Skill Lets AI Assistants Operate Desktop Apps
The author has released an open-source skill designed to let AI assistants view a screen, click buttons and type in apps. A demo shows Claude Code using a browser to find a restaurant. The tool reuses locally installed ChatGPT/Codex Computer Use components, requires Codex and does not call OpenAI’s inference API, according to the author. They say it could lower token costs while keeping app sessions and payment information on the user’s computer.
View on X (Twitter) →
367 more posts indexed for @Kay2289123 — they go out in the digest.
@Kay2289123 posted 32 times last week.
This page shows what they already posted. Get tomorrow's — and the day after — summarized in one morning email. First 7 digests free, no card required.
Follow @Kay2289123 by email →
