2025 Year In Review
December 28, 2025
As the Rolling Stones say, you sometimes don’t get what you want, but you get what you need.
2025 happened to be there year for me: a lot of accomplishments that were very different from the ones I wanted. But I did learn a lot!
The sub-4 marathon journey continues
I’ve been hard-pressed this year to try and break the four hour marathon. I did two marathons this year: the Jersey City Marathon and the Philadelphia Marathon. Both times, I got shy of breaking 4 hours.
I was pretty disappointed after the Jersey City Marathon, because I did try very hard for that one and was probably in the best running shape in my life at that point, and crashed out due to poor strategy (aka running way too fast in the beginning).
The Philadelphia Marathon, I was less trained for unfortunately, but had a better strategy and ultimately a better time because of it. There was a point during the Philadelphia Marathon where I started laughing in a good way, around the 21-22 mile mark, because I realized that I wasn’t going to break four hours, but that I tried my absolute best and I shouldn’t ever feel bad about that.
Anyway, the journey continues next year. I am not lucky enough to do another two marathons due to my sabbatical plans, but will try to come back swinging for the Philly Marathon 2026.
My first stint at management
Cocoon gave me the opportunity to try out engineering management for the first time. I definitely learned a lot through this experience - I personally didn’t really get as much satisfaction as I did from coding, but I did feel more accomplished seeing my direct reports accomplish and develop in the way they did. It was definitely a different kind of high, seeing someone develop a skillset with your intentional support over time.
I would say the pressure here is different from just being a mentor, or influencing from the side. I feel like in that kind of role, you can offer advice and nudges (which can still change the trajectory of how things are going), but you are not ultimately the arbiter of a person’s career. I’ve learned that being an EM very much means you are the arbiter in that regard, especially since you have access to more of the truth and chaos of the business.
This round of managing was hard since I was asked to balance it with my IC duties. The infamous TLM role! It was really hard being a TLM because as an IC, I get very emotionally invested in my projects, and sometimes that emotion can cloud judgement as a manager. Also, it was an extreme sap on my time - being an EM is a full-time job, and my opinion after doing it once is that anyone that is trying to be an IC or EM is probably overstretched in one area or both.
Overall, I decided after 9 months that while it was fun, being a TLM was a little hectic, and I wanted to spend more time advocating for the technology and being a full-time Staff Engineer again. I switched back to being IC for the remainder of my time at Cocoon.
AI Programming Tools come back swinging
I took a short staycation in June this year, and at that time I really tried out Claude Code for the first time, and was utterly amazed at how good it was. I didn’t really understand agentic coding before trying out Claude Code, but I’m glad I did because it really opened my eyes on what the future of programming and AI can really be.
I had some exposure to Cursor before trying out Claude Code, and while that experience was magical as well at first, it feels too much like a walled garden, and I think no matter how much the Cursor team tries to make the IDE shinier, it’s hard to escape the gravitational pull of being an IDE. (I do realize that there is a cursor-cli, but I haven’t tried that yet since it didn’t feel like Cursor’s main product.)
Versus while using Claude Code, I’ve found my imagination is almost the limit for what I can do with it as a general agentic framework. I’ve found that you can use it for more than just writing code, which is very exciting to me. I’ve used it to help keep my notes organized, with some translation tasks I’m doing for a side project I’m working on, and much more [1].
That, and Anthropic continues to be on the forefront for creating new standards for agentic coding (like with Skills). Opus 4.5 was a marvel in the last 3 weeks I got to use it for professional work at Cocoon. I’m excited to see how companies like Anthropic continue to push the boundaries of coding.
Ultimately, I know I am looking at just the tide pool of an ocean of AI programming tools that exist right now. But I’m just marveling at how in 2024, my ceiling on what AI programming was was copy/pasting from ChatGPT, versus now it’s having a full-blown mini-army of agentic interns that can run in parallel to each other.
I was middle-of-the-ground unsure about programming with AI coming into 2025, but now at the end of the year, I can firmly say that I’m in the “this is the future” camp now [2]. I can’t wait to see and experiment on any new paradigms in tools that come in 2026.
Taking the next step
Now that I’ve had more time to process my transition out of Cocoon, I’m happy to share a little more about how that decision was made. I decided after doing a backwards calculation that for what my wife and I want to do in the future, it makes sense for both of us to take an intentional sabbatical from work and spend time together, before our next phase of life. This gives us time to travel, for me to explore some of my own projects, and for me to really think about what I want to do as a next step without rushing [3].
It was a tough decision, a few months in the making, but once it was made it felt right in my gut. To me, a true gut feeling means a calm knowingness that it’s right, despite it being scary [4].
Telling my manager and skip manager wasn’t easy, but they were supportive of my decision and were excited for the next part of my journey. I’m lucky enough to have had very understanding managers at Cocoon, and I’m thankful that my exit was handled with a lot of grace.
If there’s one piece of advice here that I’ve learned, is that the goldilocks amount of time for a proper transition out is one month. Any more than that and you are just wasting people’s time and your own opportunities, any less than that could be rushed especially if you have some kind of leadership role. So if you can afford to do so and are lucky enough to have a graceful exit, shoot for letting your bosses know one month in advance.
Ultimately, I can already feel myself itching to go back, but I’m trying to not rush the next move. I’m taking time to really internalize what I need for my next role in a careful and delibrerate way. Of course, I am planning to write some extended thoughts on how I’m planning for and making this choice when it’s time.
[1] I’m very much due to writing a bigger blog post about this, but in just a sentence, I think that if you have some baseline coding skills and you have access to Claude Code, you have pretty much all the AI functionality you will ever need, coding and beyond.
[2] I am very aware of the AI bubble talk and I think one can treat the financial side of AI and the utility side of AI with different mindsets. These tools are genuinely useful, and have helped me accomplish things much faster with less cognitive overhead! But it is genuinely crazy how much money is being poured into these tools, with the hope they will replace ALL valuable forms of white collar work.
[3] This is a great article by Molly Graham that made me change my perspective on intentional breaks between jobs.
[4] Sometimes, it can be easy to confuse gut decisions with excited decisions, but excited decisions, IMO, can be more dangerous because you are not fully processing the information. You are making a decision on passion and “vibes”, and you can unfortunately reap what you sow if you’re not careful.