@yan5xu: Sharing an AI company research library I recently compiled: Oh My AI Company.

Sharing an AI company research library I recently compiled: Oh My AI Company. After bb-browser was acquired, as a co-founder, I have a very important and time-consuming task: continuously tracking market directions — where new product forms are heading, which technologies are maturing, what enterprises are truly willing to pay for, who is raising funding, who is growing, and which companies are competitors versus adjacent. In the past, this information mainly came from X, WeChat public accounts, company blogs, and daily feeds. Every time I saw an interesting company, I had to do extensive deep research myself. This approach allows deep insights, but is limited by personal time and makes it hard to continuously cover a large market scope. So I started thinking about whether an agent could take over part of this work. Gradually, this process formed a research loop: input a new company, a funding news, an investor or product clue, and the agent will expand along directions such as product, founders, funding and investment firms, technology and docs, traffic and GTM, community feedback, competitive relationships, etc. The final output is not an isolated report, but a small market map that can continue to grow. As the volume of research increased, we encountered a new problem: if these companies, people, sources, and judgments were scattered across chat logs or Markdown, it would still be difficult to reuse them in the next round. Memex was also developed during this process. What we needed was not just a note-taking tool, but an infrastructure that allows humans and agents to jointly maintain long-term research memory: using structured objects to store companies, people, and sources, using Markdown to retain readable research text, and then connecting them through graph relationships. This library also includes a type of data that is relatively rare externally: Similarweb traffic analysis. I don’t use it simply to compare which company has more traffic, but to use third-party traffic signals to assist in judgment: Whether a product has formed real access scale, or only has funding and PR buzz; Whether growth comes from SEO, advertising, community, LinkedIn, Product Hunt, or media coverage; Whether users are searching for brand names, specific needs, or competing products; Whether seemingly similar companies are direct competitors, adjacent products, or noise from overlapping audiences. For example, we initially wanted to determine whether Viktor's growth was mainly driven by funding PR. The traffic structure showed that it was more like a launch powered by PR, search, social, and paid advertising combined, and a significant amount of official site visits continued into the product app, not just content exposure. Another example: Browserbase, Hyperbrowser, and Browserless all superficially belong to browser execution, but when combining traffic sources, search terms, GitHub, Docs, and community signals, we can further distinguish developer agent infra, early developer GTM, and more mature scraping/browser automation infrastructure. They cannot be simply placed into the same competitor bucket. Of course, Similarweb is only a third-party estimate, more suitable for observing scale, structure, and trends, and cannot replace official data or first-hand user validation. Therefore, the library also cross-validates traffic data with official websites, documents, GitHub, funding reports, and community discussions. As of now, AI Company Atlas has compiled: 75 AI companies and products 75 investors or investment firms 111 founders, investors, journalists, and key practitioners 227 continuous tracking entry points Key coverage includes AI products, AI infra, enterprise Agents, Browser/Agent Infra, and vertical AI. It is not a company ranking, nor several one-time industry reports, but rather a continuously updated market map. I will continue to update it. If anyone is also working on AI products, Agents, or related research, feel free to use it directly. If you find any information errors or companies worth following up on, please let me know or directly submit an Issue / PR.