I've been at the Jules team for ~1 month, and I want to take a quick snapshot of how I feel right now:

• Googlers working at Labs are more ethusatic about programming and technology in general. After returning to work, many would share what they built during the holidays, which is unheard of when I was in Ads. Of course, this is largely due to the fact that many work on coding agent and want to have a playground to test it, but it still feels quite different.

• Teams work quite differently. My previous teams use Email + Chat + Docs evenly for communicaton, but the Jules team use Google Chat 90% of the time. This allows the team to communicate more efficiently and get in-time feedbacks.

• The Jules team (and Labs in general) moves a lot quicker than all my previous teams. This is within expectation, but it's still interesting to see new features added to the product every week, compared to taking quarters or years to launch something.

• I can finally learn what I want to learn, and more importantly, not feeling disconnected from the rest of the world. Previously, reading articles written by other AI leaders/labs feels like a pain: "ok I learned a context engineering trick, so what? It's irrelevant for me." Now, I can use what I learned on my daily work. My 9-5 and after-work life finally doesn't feel like two different worlds, I'm finally a whole person without schizophrenia (精神分裂症).

• Building agent is quite engineering heavy, and most techniques are about overcoming the limitations of LLM. e.g. LLM can't count line numbers, thus I need to prepend line numbers to diffs before injecting it into the prompt.

• Using Python, the language I'm proficient with, feels good. Yes LLM write most of the code these days, but language proficiency can help you make wise decisions, avoid pitfalls and anti-patterns.

• Pivoting seems to be a norm here. I will share more when the time comes.

Overall, how things turned out aligns with my expectations. I'm excited for what's next.