How SuperMe works
SuperMe, a startup building a network of AI profiles made from real experts’ knowledge, has raised $20 million in a round led by Kirsten Green at Forerunner, with Greylock Partners participating. Announced on 29 October 2026, the funding will scale both sides of the network: the people asking questions and the experts whose knowledge powers the answers. Instead of chatting with one general-purpose assistant, you ask a question and the platform picks the best-matched experts for it.
The result is a group chat with the AI profiles of those experts, who can discuss the problem with each other while you watch live or read a summary afterwards. You can also ask a single profile directly instead of the whole group.
As an example, a request to analyze a pitch deck could be answered by AI profiles of venture capitalists Mercedes Bent, Sarah Tavel and Josh Elman, each drawing on what that person has actually said and written in public before.
The profiles are built from public material and, where experts connect their apps, private sources too. SuperMe scans public sources such as LinkedIn, YouTube, Substack and blogs, and can also pull from connected apps including Google Drive, Notion, Gmail and Granola.
The company says some of that material may include personal information or trade secrets, but it ensures none of it appears in the answers the profiles give, a safeguard that matters because profiles draw on real people’s work and reputations.
Who built it
SuperMe was founded in 2024 by Casey Winters and Ludo Antonov, who connected while Winters was advising Whatnot. Winters has held executive roles at Grubhub, Pinterest and Eventbrite and has advised Airbnb and Canva, while Antonov was head of engineering at Pinterest, Lyft and Whatnot.
The pair launched the product publicly about a year ago, with the group-chat format and access from inside Claude and ChatGPT added more recently. Both additions matter: group chat is the product’s signature format, and chatbot access puts it where questions already happen.
The company is free to use and plans to introduce a question cap followed by a paid tier. It is considering how to reward the experts whose knowledge the network runs on, a question investor Kirsten Green flagged: the supply of expert insight, she said, needs incentives and benefits for contributors if the model is to scale fairly.
SuperMe has around 13 employees split between San Francisco and Ho Chi Minh City, and it is hiring for conversational-AI and matching roles plus an expert-scouting role. That hiring pattern shows where the hard work sits: not just in generating answers, but in matching the right expert to each question and finding experts worth adding next.
Why It Matters
SuperMe sits in a wave of personal AI agents that act on your behalf rather than answering prompts. It works through the Model Context Protocol, so users can ask its expert profiles from inside Claude and ChatGPT rather than opening an app. A similar direction is visible in Ant’s LingGuang 2.0 personal agent, which acts without waiting to be asked.
The bet is that expertise, not intelligence, becomes the differentiator. A general chatbot can summarize a topic, but an answer shaped by a specific investor’s thinking, or a founder’s operating experience, is harder to copy. That is the same logic behind Wabi’s $55 million personal AI messenger, which builds its assistant around one user’s life rather than one expert’s knowledge.
The questions are accuracy, consent and incentives. A profile built from writing is an interpretation of an expert, not the expert, and users may treat its answers as advice. If SuperMe can keep profiles faithful, pay experts fairly and stay inside the chatbots where people ask questions, it has a path; if any slips, the network’s trust advantage disappears.
Background
Forerunner’s Kirsten Green led the round, with Greylock Partners also investing. SuperMe launched publicly about a year before the October 2026 announcement, and the group-chat format plus Claude and ChatGPT access are among its newer additions.
Source: TechCrunch report.
0 Comments