I just got back from LA after spending Christmas with my wife’s family and friends. We did a TikTok trend called the “accomplishment cake,” which gave me a chance to reflect on 2025. The trend is meant to be positive, but during the 7-hour drive home, I also spent some time thinking about the negative.
Work
2025 humbled me a lot. Compared to 2024, when I got promoted at work, got accepted into YC, and even had a Times Square billboard for Palmier, 2025 has been a lot less flashy and much more grindy. We’ve pivoted twice this year, and we’re still trying to figure out what problems we’re solving and how we’re solving them.
By the tangible metrics, we didn't have good progress and had a rough year. low DAUs, low revenues, pivoted. My mental health got a hit (dw just a small dent), and all I could think of was "just don't die".
By the intagible metrics, I've grown a lot as a builder, and would like to celebrate the small wins and learnings.
Palmier Coding Agent
We had around 400 users, and I learned a ton while building it:
- MCP auth – This was a lot to deal with, but we ended up building a solid system where you can one-click OAuth the MCP for a coding agent that doesn’t run on your laptop.
- Context engineering – Like most other coding agents, context is everything.
- Workflows – We started with just a queue and a worker, then realised we needed durability and retries, so we moved to Temporal. Later, we realised it was too expensive and complicated for our use cases, so we migrated to Hatchet and became very happy users.
- Research – Before launching, we did a lot of research on different approaches, like agentless, LightRAG, and SWE-agent. We ended up using a simple while loop with hooks.
- Slack / GitHub bots – Some simple client integrations to our server. This really made me realise that integration is king.
- Frontend – The above was human-written with AI assistance. Our frontend, however, was mostly AI-written with human review, mainly because we lack frontend experience prior this.
toktop
This was a weekend project I built to solve my own problem: constantly checking my OpenAI and Anthropic usage on two different websites, both of which kept logging me out. Toktop is a TUI that takes your local admin keys and lets you view your cost and usage directly in the terminal, where I already spend most of my time with Claude Code.
I deliberately chose Rust as part of my learning. This is also my first personal open-source project, so I decided to promote it on Discord and Reddit. It ended up getting 145 GitHub stars and around 300 downloads.
sixsevenstudio
Around September, Sora 2 came out, and not many people had access to it. I built a desktop app that let you generate Sora videos using an OpenAI key (I later realized non-developers don’t really use OpenAI keys… oops). The idea was to generate a story → images → videos with just one API key.
I built this mainly as a learning project and never planned to monetize it. It’s no longer maintained since I’m focusing on the cloud version at Palmier.
My personal blog
Writing a simple static website is incredibly easy now with tools like Claude Code and Cursor, so I decided to maintain this blog to improve my writing and design skills. You can build great websites in a short amount of time, but having good design sense as a developer is still hard. This site will probably change a lot as I spend more time working on design.
Personal
I’ve had a great year. I proposed to and married my wife. We got a second cat named Maui, at the same place where I proposed. I surfed more over the summer and got better at it. I ran an unofficial half marathon by myself (1:54). I benched 225 lbs. I went home to see family and friends for the first time since 2019—many of whom I hadn’t seen since high school. I’m glad they still remember me (a rare positive outcome of social media).
Honestly, I’m grateful for all of it. Despite the setbacks and challenges, I’m living a good life.
2026
I keep reminding myself that "progress is not linear", and I’m genuinely excited for 2026. My resolutions for the year:
- Mental health – This will be my biggest priority. Startups are hard, and no one can make it without taking care of themselves. I used to not understand why people meditate, but I want to try it and see how it feels.
- Physical health – I want to bench more, with a goal of 275 lbs. I’d also love to surf more when it gets warmer and get back into running, which I gave up halfway in favour of gym sessions.
- Coffee – I’ve been posting my lattes on Instagram. My goal is to get better at latte art and eventually be able to draw a swan.
Happy new year!