Cosmos Institute granted Zhongying Qiao on August 15, 2025 for Socratic Mirror, with funding in the $1K to $10K range plus compute. The grant came through the inaugural AI x Truth-Seeking cohort of 27 winners co-funded by Cosmos Institute and FIRE under a $1M program.
Qiao is a senior software engineer on GitHub's security products team and an author on The GitHub Blog covering AI and machine learning topics. He holds joint master's degrees in computer science from TU Berlin and the University of Trento, and a bachelor's in electronics and electrical engineering from the University of Glasgow. Prior roles include PhD research assistant at the University of Notre Dame (IT security, networking systems, big data) and research at Fraunhofer IPK on machine-learning techniques for image and video analysis.
Socratic Mirror is positioned as an AI thought partner designed to interrogate the user rather than answer them, modeling Socratic questioning to push deeper thinking instead of optimizing for user satisfaction. The product framing inverts the default assistant pattern, in which a system competes to give the most agreeable answer. The cohort's 90-day prototype window applies, so the deliverable target is a working demo of the inverted-assistant pattern rather than a production product.
Within the cohort Socratic Mirror sits in the agents and evaluation cross-section, in a thematic cluster with Socratica and Argument Debugger, all three of which treat the quality of the user's reasoning as the unit of value rather than the helpfulness of the assistant's answer. The lock-in vector being addressed is the bias in instruction-tuned assistants toward sycophancy and engagement, which the cohort's framing treats as a distortion of the user's own inquiry.
Recipient
Zhongying Qiao
Funder
Cosmos Institute · foundation · US
Backs philosopher-builders making prototypes, essays, and projects at the intersection of AI and human flourishing, with emphasis on reason, decentralization, and individual autonomy.
Primary source
https://blog.cosmos-institute.org/p/introducing-the-first-cohort-of-ai
Additional sources
More from Cosmos Institute
- Institute for Decentralized AI launch 2025-09 · 5 fully funded research positions
New institute combining research and infrastructure to support AI that operates without centralized gatekeepers. Initial cohort of five fully funded positions across Oxford and Stanford.
- Metalens 2025-08-15 · $1K-$10K plus compute
Open-source scientific research platform combining AI summarization with structured human checks, intended to help researchers see beyond hype and identify reliable claims in dense literature.
- Perplex 2025-08-15 · $1K-$10K plus compute
Tooling for surfacing hidden goals in closed AI systems by probing them with open-weight reference models. Aims to give independent reviewers a way to audit proprietary deployments.
- Policy Explorer 2025-08-15 · $1K-$10K plus compute
AI tool for analyzing the impacts and assumptions embedded in regulatory and policy proposals, geared toward helping civil-society analysts surface what a rule actually does.
- Argument Debugger 2025-08-15 · $1K-$10K plus compute
AI assistant that finds gaps in chains of reasoning and suggests repairs, framed as a tool for better arguments rather than persuasive ones. Built by a Harvard CS faculty member.