index dfeb322..1e9f671 100644
@@ -4,116 +4,138 @@ visibility: public
# competitions & hackathons
-competitions are one of the few things where being young is genuinely an advantage. the bar for "impressive" is lower, the communities are tighter, and the skills transfer directly to real work. but not all competitions are created equal.
+i've done a lot of competitions. here's an honest account of every one — what it was, how I did, and what I actually learned.
-## math competitions
+## hackathons (3.5 wins out of 5)
-math competitions teach you to think precisely under pressure. the skills transfer to everything — CS, research, startups, writing. the people you meet are also disproportionately interesting.
+hackathons are where I learn fastest. you compress months of learning into 24-48 hours, and you come out with something real.
-### AMC / AIME / USAMO pipeline
-- **what:** AMC 10/12 → AIME (top ~5%) → USAMO (top ~250) → IMO team (top 6)
-- **honest take:** the canonical math competition pipeline. genuinely useful for building mathematical maturity. but it's also the most crowded lane — tens of thousands of students grind AMC problems for years. if you enjoy it, great. if you're doing it because you think you need it for MIT, reconsider. making AIME is meaningful. USAMO is elite. but "I scored 100 on AMC 12" is not a personality.
-- **time commitment:** moderate for AMC, extreme for USAMO
+### TRAE Solo Hackathon (Oct 2025)
+- **result:** 1st place out of 29 ($1500)
+- **where:** Roblox HQ
+- **what I learned:** solo hackathons force you to scope ruthlessly. no teammate to split work with means every feature choice matters.
-### HiMCM (High School Mathematical Contest in Modeling)
-- **what:** 36-hour team math modeling competition. you get a real-world problem and build a mathematical model.
-- **honest take:** massively underrated. this is the closest competition to actual applied math/engineering work. you learn to model, code, write reports, and work under pressure. the problems are genuinely interesting. teams of up to 4. I've done this and highly recommend it.
-- **when:** november
+### AGI House Agent Skills Build Day (Mar 14, 2026)
+- **result:** won
+- **team:** built a Claude skill with Adele and Sefika. also launched referral.bike at the same event.
+- **what I learned:** the best hackathon teams have complementary skills, not overlapping ones.
-### MCM/ICM (Mathematical Contest in Modeling)
-- **what:** the college version of HiMCM. 4-day competition, open to undergrads (and technically high schoolers).
-- **honest take:** harder and more prestigious than HiMCM. if you're a strong team, competing as high schoolers is a flex. the problems are fascinating — real interdisciplinary modeling challenges.
-- **when:** february
+### EventConnect at Manus AI event
+- **result:** won audience vote
+- **team:** built with Dirac
+- **what I learned:** audience votes reward demos that feel magical. polish the demo.
-### M3 Challenge (MathWorks Math Modeling Challenge)
-- **what:** 14-hour math modeling challenge. free to enter. teams of 3-5 high schoolers.
-- **honest take:** similar to HiMCM but shorter and free. good entry point into math modeling competitions. prizes are decent ($100k+ total).
-- **when:** march
+### MongoDB Agentic Memory & Context Engineering Hackathon (Oct 11, 2025)
+- **result:** top 6 out of 70
+- **where:** Cerebral Valley
+- **what I learned:** "top 6 out of 70" feels like almost-winning, which is its own kind of useful. close losses sharpen you.
-### MTFC
-- **what:** team math competition
-- **honest take:** less well-known but good for building team math skills without the AMC grind pressure.
+### Gemini Multimodal Hackathon (Oct 18-19, 2025)
+- **result:** produced OnCue demo
+- **what I learned:** multimodal demos are hard to scope. the "wow" moment needs to be in the first 30 seconds.
-### USAYPT (US Association of Young Physicists Tournaments)
-- **what:** team physics tournament. you prepare research on assigned problems and debate other teams.
-- **honest take:** unique format — it's physics meets debate. you present your research, then other teams try to poke holes in it. teaches you to defend your ideas rigorously. very different from "solve this problem in 30 seconds" competitions.
+### organizing: Startup Pitch Hackathon (May 9, 2026)
+- organizing this with Doreen Xia. $5k/$2.5k/$1.25k prizes. being on the other side of hackathons teaches you what judges actually look for.
-## science fairs & research competitions
+### upcoming: SF Hackathon (Apr 25, 2026)
-### ISEF (International Science and Engineering Fair)
-- **what:** the biggest science fair in the world. ~1800 finalists from 80+ countries.
-- **honest take:** ISEF is legitimate. getting there requires winning a regional/state affiliated fair first. the research quality at the top is genuinely impressive. but the pipeline (regional → state → ISEF) is long and political — your local fair's judging quality varies wildly. the biggest trap: spending a year on a project that's optimized for fair judges rather than for being good science. do the research because it's interesting, then submit it to fairs. don't do research because you want to win a fair.
-- **the ISEF meta:** projects with disease/cancer in the title disproportionately win. machine learning projects are now so common that judges are skeptical. novel hardware, environmental science, and behavioral science projects are currently undervalued.
+### hackathon advice (earned from actually winning)
-### Regeneron STS (Science Talent Search)
-- **what:** the oldest and most prestigious science competition in the US. submit a research paper. 300 scholars, 40 finalists, top prize $250k.
-- **honest take:** the research quality required is high — this is basically "did you do publishable research?" judges are actual scientists. less gaming potential than ISEF. if you've done real research, apply. if you haven't, this isn't the place to start.
+1. **the demo is everything.** judges spend 2-5 minutes with your project. if you can't show something impressive in that time, nothing else matters.
+2. **scope aggressively.** the #1 mistake is building too much. pick one impressive thing and polish it.
+3. **the 70/30 rule:** spend 70% of your time on the core feature and 30% on the demo/presentation. most teams invert this.
+4. **pick the right prize track.** "best use of [sponsor API]" categories are usually less competitive than "best overall." using a sponsor's API well is often a free win.
+5. **tell a story.** judges remember the team that had a compelling "why" more than the team with the most features.
+6. **ship it.** having a live URL or working app beats a slide deck every time.
-### JSHS (Junior Science and Humanities Symposium)
-- **what:** present original research. regional → national. DoD-sponsored.
-- **honest take:** lower profile than ISEF/STS but solid. the presentation component means you also learn to communicate your work. good stepping stone.
+## math competitions
-### Synopsys Science Fair (now Broadcom MASTERS for younger students)
-- **what:** regional science fair, often a feeder to ISEF.
-- **honest take:** the quality varies by region. Santa Clara County's is one of the best in the country if you're in the Bay Area.
+math competitions teach you to think precisely under pressure. the skills transfer to everything.
-## hackathons
+### AMC 10
+- **scores:** 91.5/73.5 (2024), 88.5/97.5 (2023)
+- **honest take:** the canonical pipeline. I did it, it was useful for building mathematical maturity, but I didn't make AIME and that's fine. the AMC grind has diminishing returns if you're not naturally headed toward USAMO.
-hackathons are where builders actually learn to build. they compress months of learning into 24-48 hours.
+### HiMCM (Nov 2025)
+- **team:** Dirac, Parth, Daniel
+- **what we did:** fire/evacuation scenario, Monte Carlo simulation
+- **honest take:** massively underrated competition. this is the closest to actual applied math/engineering work. 36 hours, real-world problem, build a model. I love math modeling competitions more than contest math.
-### where to find them
-- **[Major League Hacking (MLH)](https://mlh.io):** the official organizer/sanctioner of most hackathons. their event calendar is the best starting point.
-- **[Devpost](https://devpost.com):** where hackathon submissions go. browse past winners for inspiration.
-- **Hack Club's hackathon list:** community-maintained list of high-school-friendly hackathons.
-- **your local scene:** the Bay Area, NYC, and major college towns have hackathons almost every weekend. find the local organizers on Twitter/Discord.
+### COMAP MCM/ICM
+- **result:** Meritorious (top 10%)
+- **what we did:** spectral bisection, cellular automata, graph traversal
+- **honest take:** harder and more prestigious than HiMCM. competing as high schoolers against college teams and placing top 10% was a genuine accomplishment.
-### how to actually win
-having won 3.5 out of 5 hackathons, here's what I've learned:
+### M3 Challenge
+- **result:** 143/770 (top 19.8%), qualified for second round
+- **honest take:** 14-hour math modeling challenge, free to enter. good entry point into modeling competitions. we qualified for the second round which felt great.
-1. **the demo is everything.** judges spend 2-5 minutes with your project. if you can't show something impressive in that time, nothing else matters.
-2. **scope aggressively.** the #1 mistake is building too much. pick one impressive thing and polish it.
-3. **the 70/30 rule:** spend 70% of your time on the core feature and 30% on the demo/presentation. most teams invert this.
-4. **pick the right prize track.** if there's a "best use of [sponsor API]" category, it's usually less competitive than "best overall." using a sponsor's API well is often a free win.
-5. **tell a story.** judges remember the team that had a compelling "why" more than the team with the most features.
-6. **ship it.** having a live URL or working app beats a slide deck every time.
+### MTFC
+- **result:** semifinalist (2025-26)
+- **what we did:** equitable bus routing
+- **honest take:** the problem was genuinely interesting. applied math with social impact.
+
+### IMMC 2026 (Mar 18-23)
+- international math modeling. similar flavor to HiMCM/MCM but with an international pool.
+
+### BmMT
+- **result:** 2nd best puzzle round score (2023)
+
+### SMT (Stanford Math Tournament)
+- **result:** honorable mention individual + team
-### the hackathon-to-startup pipeline
-some of the best projects start at hackathons. if you build something people want during a hackathon, keep going. the hackathon is the prototype; the startup is the product.
+### NMT (Nueva Math Tournament)
+- I co-lead this. organizing a math tournament teaches you logistics, problem-writing, and community-building all at once.
-## business competitions
+## physics competitions
-### DECA
-- **what:** business roleplay competitions. regional → state → ICDC.
-- **honest take:** DECA teaches you to think on your feet and communicate business ideas. the roleplay format is weirdly useful for founder skills — you're basically pitching under pressure. the actual business knowledge is surface-level, but the soft skills are real.
+### USAYPT (Jan 2026)
+- **result:** 2nd nationally
+- **team:** Harrison, Parth, Mo, Eddy, Clara, Colin, Emily, Aurelia, Liam, Kevin, Andy
+- **honest take:** unique format — physics meets debate. you present research, other teams poke holes in it. teaches you to defend ideas rigorously. placing 2nd nationally with this team was one of my proudest moments.
-### FBLA
-- **what:** similar to DECA but broader. events range from coding to public speaking to accounting.
-- **honest take:** more variety than DECA. the tech events (coding, web design) are easy wins for people who can actually code.
+### F=ma Exam (2026)
+- the qualifier for USAPhO. took it.
-### Diamond Challenge
-- **what:** high school entrepreneurship competition by University of Delaware. submit a business plan.
-- **honest take:** one of the better business plan competitions because it's specifically for high schoolers. the finals include pitching to real investors.
+## science & research competitions
-## special awards & fellowships
+### BL4S / CERN BeamLine for Schools (submitted Mar 2026)
+- **team:** Gamma Guardians — Harrison, Noah, Sydney, Abhi
+- **what:** proton beam shielding simulation
+- **honest take:** writing a proposal for actual CERN beam time forces you to think like a real physicist. even if we don't get selected, the proposal-writing process was incredibly valuable.
-### Davidson Fellows
-- **what:** $10k, $25k, or $50k scholarships for significant work in STEM, literature, music, philosophy, or "outside the box."
-- **eligibility:** must be 18 or under, US citizen/resident
-- **honest take:** this rewards depth over breadth. you need to have done something genuinely significant — published research, a meaningful invention, a substantial creative work. the "outside the box" category is interesting for builders who don't fit neatly into STEM or arts.
+### Davidson Fellows (submitted Feb 2026)
+- **what:** submitted paper #0352 on anesthetics/EEG
+- **honest take:** this rewards depth. you need to have done something genuinely significant. I submitted my EEG research.
-### Thiel Fellowship
-- **what:** $200k over 2 years to drop out of college and build something. 22 or younger.
-- **honest take:** covered more in the [funding page](/wiki/funding-grants). the most famous fellowship for young builders. extremely selective (~15 per year). you don't have to drop out to apply — they want to see you're building something that college would slow down.
+### Golden Gate STEM Fair (early Mar 2026)
+- regional science fair.
### Breakthrough Junior Challenge
-- **what:** create a 2-minute video explaining a concept in physics, math, or life sciences. $250k college scholarship + $50k for your teacher + $100k science lab for your school.
-- **honest take:** this is a video competition, not a research competition. production quality matters as much as scientific accuracy. if you're good at explaining things and making videos, this is unusually high-EV for the time investment.
-- **when:** submissions due ~september annually
+- **result:** top 40%
+- **what I did:** RL intuition video
+- **honest take:** this is a video competition, not a research competition. production quality matters as much as scientific accuracy. top 40% isn't a win, but making the video taught me about science communication.
-## the meta
+### Congressional App Challenge
+- **what I submitted:** Pause app
+- **result:** congressional certification
+- **honest take:** relatively low bar compared to other competitions, but the certification from your congressperson is a nice touch.
+
+### other competitions I've done
+- **BIG Idea Competition:** Honorable Mention
+- **Scholastic Writing Awards:** submitted Dec 2025
+- **NACLO:** open round (computational linguistics — surprisingly fun)
+- **PicoCTF 2024:** cybersecurity CTF. good for learning security basics.
+- **Wharton Data Science:** participated
+- **Fed Challenge (NY Fed):** did this with Alexis. made a corn podcast. yes, a corn podcast.
+- **Ethics Bowl:** went 2-1
+
+## chess
-competitions are tools, not goals. the best reason to do a competition is because the preparation makes you better at something you care about. math competitions make you a better thinker. hackathons make you a better builder. science fairs force you to do deep research.
+USCF 1386. won USATW U1000 6-0, 3rd at SF Scholastic, perfect score at D-Tech, bughouse 8-0. chess is the thing I do for fun that happens to also be competitive.
+
+## the meta
-the worst reason is "this will look good on my college application." admissions officers can tell the difference between someone who competed because they loved it and someone who competed for the resume line.
+competitions are tools, not goals. I did a lot of them — probably too many. the ones that taught me the most were the ones where I was genuinely interested in the problem (math modeling, USAYPT, hackathons), not the ones where I was chasing a result.
-if you're going to compete, go deep in one or two areas rather than shallow in five. a national-level result in one thing is worth more than mediocre participation in everything.
\ No newline at end of file
+if you're going to compete, go deep in one or two areas rather than shallow in five. I went too wide sometimes. a national-level result in one thing is worth more than mediocre participation in everything.
\ No newline at end of file