In July I flew to Toronto for the ICIC conference and gave an oral presentation. Between sessions I met a PhD student from UC Berkeley who showed me the potential of auto research — and it prompted me to revisit an idea I had shelved a long time ago. The topic sits in post-quantum cryptography: I wanted to test a novel attack vector, not a conventional injection, but starting from the Agent API surface. As the experiments progressed, I found results that genuinely excited me.

August was a month back home in China — spending time with family while pushing the paper forward, and I managed to make the USENIX Cycle 1 deadline. Using Claude and Codex from within China was far from smooth, but my server in Germany saved the day. I made full use of my Claude Code quota during the promotional period. There was also a minor hiccup at submission — I forgot to include the artifact URL. Fortunately, after reaching out to the chair, I was able to attach it via a comment.

I recently returned to Germany and have started applying for PhD positions, while also exploring industry AI opportunities. On the research side, I plan to keep pushing in two directions: post-quantum cryptography and Agent security. I have a couple of ideas that still need further validation. On the fitness front, I’ve kept up a push-pull-legs routine for over a month now, starting from when I was back in China, and I can clearly feel the strength and overall condition improving — a good trajectory.

Back in Germany, there’s no shortage of plans: beyond advancing my research, I’m preparing for English and French language exams. After a brainstorming session yesterday, I decided to enter some startup competitions and training programs, combining them with my research. Life keeps moving forward, slowly but surely. I don’t know — and I look forward to — where I’ll be and what I’ll be doing a year from now. Turning 32 like this feels worth looking forward to.