What I'm doing now

A sporadically updated log of what I am doing, thinking, and exploring.

Jul 9, 2026

AI safety research internship at Lida Safety Research, working on formal verification automation for vibecoding and AI governance simulations.
Co-leading the analysis subteam for a multicultural riddles benchmarking project at Cohere Labs Open Science Community, focusing on factual accuracy, hallucination behavior, topic-level patterns, and cross-cultural reasoning failures.
Finalizing scaling behavior of chain-of-thought monitoring in LLMs from Algoverse AI Research.
Contributing to MatrAIx’s collection of projects, especially harbor-based runtime environments.

Timeline

Jun 4, 2026
Started interning at Lida Safety Research, contributing to projects in mech interp and sandbagging.
May 1, 2026
Started co-leading Cohere Labs Open Science Community’s multi-cultural riddles benchmarking project with other six leads, working with 100+ community researchers.
Apr 15, 2026
Started ElectricSheep’s Futurekind AI Fellowship Spring 2026 where I research and build solutions in the intersection of AI and animal protection.
Apr 6, 2026
CommonLID got accepted to ACL 2026 Main Conference.
Feb 1, 2026
Started extending Tau2-Bench for low-resource languages as a mentee in the SEACrowd 2026 Apprentice Program.
Dec 1, 2025
Attended NeurIPS 2025 in San Diego, USA from Dec 2 to 8. Highlights: UK AISI Social, met my mentor & SEACrowd team, and talked to a few working on responsible and pluralistic AI.
Oct 1, 2025
Started working on scaling behavior of chain-of-thought monitoring in white-box models with mentor Yeonwoo (Eyon) Jang at Algoverse AI Research.
Aug 17, 2025
Joined SEACrowd core team, building website, handling internal admin, and contributing to research initiatives, including the development of vision-language model with cultural/regional understanding.

Learn more about now pages and browse others at nownownow.com