2024 Wins

I wanted to record my wins for 2024 since I often get caught up in what others think of me or my own expectations for myself. This page is for my inner child 😊 and also inner attention hoe

June §

• Reproduced SWE-bench from scratch. (although copied heavily from their repo)
• Decided to go on less dates
• Went to Korea for a friend’s wedding. Enjoyed their special day and the lovely cafes throughout Seoul. Got a facial and lots of skincare/makeup to bring back
• Setting my telic goals for 2024. It feels like this time around I’ll actually stick with it. My goals as of June 5, 2024 are a 209 day streak to:
  • meditate 1hr/day
  • exercise 1hr/day. walking or stretching count
  • spaced repetition cards complete daily
  • no more shows/movies, unless it’s a group hangout
  • some other ones I won’t share here

May §

• Decided to commit to this project on programming evals heavily inspired by SWE-bench, for the next 2 months
• Made more songs
• Went on more dates
• Meditation streak continues, and upping it :) Been ~7 days in a row meditating ~20 minutes within the first hour of waking up. I hope to keep it up.
• Have continued going weekly to qi gong.
• Experiments complete
  • 12. Claude vs. GPT on GSM8K math benchmarks - How do they compare?
• Proposals - tweaked most of them

Some days I wake up with a lot of anxiety and uncertainty, but there are gradually more days per month that are painted with love, delight, bemusement, play. Someone said to me that I felt this way because I was trying hard to make my life perfect and…when I sat with the statement, I felt some hard knot dissolved inside of me.

April §

• Launched Viva’s desktop application 🌱
• Performed a musical about Silicon Valley to ~100 friends 🎶
  • Listen on Spotify
  • This was shockingly unplanned but so so incredibly fun and it was absolutely disgusting how during the performance I was filled with so much gratitude and love for my community
• Hosted a hackathon at MIT
• Meditation streak continues since Jan 1 where I meditate for at least a minute a day. I’ve tried multiple types of meditation, and it turns out that loving-kindness meditation (yes it’s a real thing) has been the one that’s stuck for me.
• Started picking up qi gong again
• Went on a few dates
• Experiments complete
  • 8. What does a good translation interface look like for same language “translation” across contexts? 🔴
  • 9. How might a set of AI models improve code generation through iterative code and test set generation? 🟢
  • 10. GPT vs. Gemini - What does a minimalistic user interface for model comparison look like? 🟢
  • 11. What does a useful parking ticket payer agent look like? 🔴
• Proposals
  • What does a better knowledge graph structure for enhancing human learning alongside LLMs look like?
  • What does a configurable format look like to represent essential data for a single person?
  • What does a configurable format look like to represent essential data for a single person?
  • “LLM Operating System”
  • Updated all the other ones too

I’ve decided I want to complete 100 experiments and their write-ups by the end of the year. Just got to get those reps in, ya know. This means (90 exp / 8 months) = average of 11.25 experiments / month.

This month has been crazy productive. On top of that I’ve been feeling so much energy.

I am fascinated by how much more productive I’ve been on certain things like programming projects, when my cup is overflowing in other areas like music. Activities B, C, D taking up more time seems to give more energy to do Activity A. I think this is how I felt in high school, running between classes and chorus and fencing (this was another era). It was a lot but they seemed to balance each other out. In college, that balance was lost / less externally reinforced, which led to a massive mental and physical rut for years. It feels like 8 years later, I’m finally regaining control of my mind and body.

March §

• I chatted with someone who runs a civil non-profit and she asked me if she could use my graph website for her non-profit work! I was so honored. I said yes, and directed her to the source code on the justanexperiment repo
• Went to Machu Picchu. I loved staying in Cusco, Peru. There is a tremendously healing energy in the Sacred Valley.
• Viva Translate continues to be running well
• Concluded my brief contribution on a Common Core dataset at Stanford. Excited to see the final paper results
• Attended a hackathon hosted by South Park Commons and OpenAI working on a note-taking interface. Turns out you can’t really make a whole comprehensive note-taking interface in 2 days
• Signed up for a marathon 💀
• Experiment write-ups
  • 7. How is the writing quality of GPT-4 + RAG for querying over my Obsidian notes? 🔴
• Proposals
  • Back-translation for Alignment of LLM Generation of Code, Tests, and more
  • PROPOSAL Knowledge base scaffolding for humans and computers
  • and others

February §

• 1 additional experiment + write-up and a few proposal drafts
  • 6. What is it like to convert a Next.js app to PWA? 🟢
  • Proposals
    • PROPOSAL Multicomponent Systems that Self-Organize with an Information Stream
    • Proposal - Crowdsourcing a universal browser automation task graph with continuously running tests
    • Solve math theorems using graphs
• Revamped https://justanexperiment.com so it’s simpler and something I would actually use
  • Translate page
  • Graph page
• Did some consulting for MultiON, which is a multi-agent startup. Their product is SO cool. It took up more time than expected but it was very fun :)
• Got my work for this Common Core research at Stanford to work fairly well.
• Released first song ever on Spotify! It is called Group House, and is about living in the San Francisco group house scene. It is a meme
• Did a hackathon hosted by OpenAI and South Park Commons. We worked on connecting graph notes together. I learned a lot about new full-stack tools like Astro
• Went on a hike with my parents and family friends. I got separated at some point and decided to play around memory techniques.

I spent a lot of time this month catching up with family I’d been neglecting during the startup journey in the last few years. I lived with my parents in SoCal, celebrated Chinese New Year with my relatives in Los Angeles, and caught up with my cousin in San Francisco. I am always shocked how similar I am to my relatives. Our speech patterns, our cheekbones, our boisterous laughter. Patterns of guilt and shame. Body language of showing affection. Maybe it is surprising to me because I spend so much time away from family, working with people whose cultural and geographical backgrounds are so different from mine. My chest feels warm when I think of them.

January §

So much happened this month!

• 5 experiments + write-ups complete and 14 additional proposals for future experiments
  • 5. What does a suite of personal GPTs on ChatGPT look like for content like research proposals and social media posts? 🟠
  • 4. What does it look like to automating UI Improvements with GPT-Vision? 🟢
  • 3. How accurate is RAG + GPT-4 for an orthopedic surgeon? 🟢
  • 2. What would a knowledge graph of science look like, generated recursively entirely by GPT-4? 🟢
  • 1. How do I track my hours in a way that feels good? 🟠
  • 0. Will I be able to build this site in 1 day, where I’ll stick with it? 🔴
• Meditated everyday 🕊️🕊️🕊️
• Used LlamaIndex for the first time to help my brother build a RAG application for Orthopedics
• Experimented with Midjourney and created a cute Insta influencer with orange hair (5 followers :D )
• Interviewed at a big tech company and a seed-stage AI startup. I feel honored that they were both direct referrals that I didn’t seek out, and that they were interested in me 😌
• Won 2nd place at a hackathon for a heatmap-to-frontend-code tool
• Assembled a few knowledge graphs
• Had a temporary obsession with Vercel’s developer ecosystem. I was able to get this multi-tenant application up and running in an hour with one of their templates. Still not totally over it yet.
• My startup Viva Translate is running well. They have their Microsoft Teams integration in development, plus some very cool generative AI features. So grateful for this team. Maybe we should rename ourselves to Universal Translator…
• Conducted an analysis over my Obsidian notes using LlamaIndex. Couldn’t find much use for Q&A flow with my notes yet, unfortunately
• Created 4 custom GPTs using OpenAI’s consumer interface. I chat with them about half with voice and half with text. I use them to create proposals and online posts - very helpful!
• Attended a lovely gathering at Replit headquarters called Playspace. I was pleasantly surprised to see so many friends that I’ve gotten to know over the past 2 years post-pandemic
• Started working part-time with AGIHouse to help build their open source AI applications and documentation
• Started research part-time at Stanford on reliably generating Common Core math problems
• Cool and very fun collaboration with some folks on agentic systems
• Used GPT-4 + Whisper in OpenAI’s consumer app to review this month’s reflection, and wow it was better than my last therapist at giving advice lol. Completely out-of-the-box too! (no system prompt)

🥵 I remind myself daily to slow down, because I feel my tendency is to move too fast. I suppose when you’re having a good time, you can get wild amounts of stuff done! There are also a lot of open threads too, and I remind myself that it’s ok to be in the ‘explore’ phase, as long as I am fulfilling my commitments to others.