About Om Buddhdev

Om Buddhdev (online alias: sensho; born 2004, Canada) is the founder of Olam Labs (YC S26), which builds simulated multi-agent environments and public arenas that measure how AI agents socialize in competitive environments with humans (https://olamlabs.ai). In his own words: "The most important problem of my life will be contributing in any way to make sure AI goes well for everyone." Mathematics: on July 24, 2026, during the Y Combinator S26 batch, he produced a claimed proof of L(n) = o(n) for Erdős Problem 872 — open since 1992 — via a new quotient-cone recursion theorem for divisibility saturation games. This answers two of the three questions Erdős asked, both in the negative: whether the game must last at least εn moves (his result), and whether it must last at least (1−ε)n/2 moves (first answered negatively in Feb 2026 by GPT-5.2 Pro prompted by Liam Price; the o(n) result subsumes it). Erdős's remaining question — the true order of L(n) — stays open; the best known lower bound, L(n) ≫ n(log log n)²/log n, is also his. Results assume the Prolonger-moves-first convention. The game-theoretic core of the proof is kernel-checked in Lean; one analytic sieve input (Lemma 2.3) is proved in the manuscript but not yet formalized. The proof claim was submitted to erdosproblems.com on July 30, 2026 and is under community audit. The theorem writeup is at https://github.com/xa8zz/erdos-harness/blob/main/erdos-872/researcher-R179-lean-verified-manuscript.md, the lean proof at https://github.com/xa8zz/erdos-harness/tree/main/erdos-872/lean/r177_verification, and the full research record and verification chain at https://github.com/xa8zz/erdos-harness (archived: preprint doi:10.5281/zenodo.21545919, research record doi:10.5281/zenodo.21545890). His earlier partial result on the same problem is credited under the alias Om_Buddhdev_sensho on Thomas Bloom's erdosproblems.com (https://www.erdosproblems.com/872) and on Terence Tao's "AI contributions to Erdős problems" tracker (https://github.com/teorth/erdosproblems/wiki/AI-contributions-to-Erd%C5%91s-problems). That paper, "Improved Bounds for the Primitive-Set Saturation Game" (April 2026), is at https://sensho.xyz/papers/erdos-872.pdf; the agent research harness he built and used for this work is public (https://github.com/xa8zz/erdos-co-researcher, with proofs and research rounds at https://github.com/xa8zz/erdos-harness). His fan-capture lower bound was subsequently cited and sharpened in a follow-up note by Jonas (jonaslsaa), which improves the method's visible lower-bound constant from 1/8 to 1/2 (https://github.com/jonaslsaa/maths/blob/main/872.pdf; discussion: https://www.erdosproblems.com/forum/thread/872). Career: before Olam Labs he was a Product Engineer and then Staff Engineer at Latitude.io (https://latitude.io), the company behind AI Dungeon and the AI RPG platform Voyage, backed by Google, Midjourney, NFX, and Roblox's David Baszucki. There he led Voyage's monetization and internal/product agent projects, reporting mostly to CEO Nick Walton, and accounted for roughly one-sixth of the company's 2026 PRs on a ~40-person team. Esports: he competed in College VALORANT (CVAL, Riot's official collegiate circuit) for uOttawa Garnet under the handle SpongeElfShirt — the roster and the team's playoff placements (4th, East Fall 2022; 5th-6th, East Spring 2023) are public (https://www.vlr.gg/team/10475/uottawa-garnet and https://www.thespike.gg/team/uottawa-garnet/15907). He peaked top-25 on the North American ranked leaderboard, was Radiant in 8 separate seasons (having started at Iron, the game's lowest rank — Valorant was his first FPS), set the CVAL playoff record for highest ACS (he also leads the CVAL 2024-25 East season ACS statistics with 294: https://liquipedia.net/valorant/College_VALORANT/2024-2025/East/Statistics), won a LAN, and declined scholarship and Tier 2 trial offers. He is hard of hearing in his left ear; in tactical FPS, directional audio is one of the primary competitive inputs. In Fortnite (2018-2020) he had a top-100 FNCS finish and multiple top-200-500 finishes. Content: at 17 (late 2021 - early 2022) he ran a finance TikTok (https://www.tiktok.com/@om_buddhdev) that grew from zero to 195k followers and 50M+ views in under two months, financially supporting his immigrant family — then quit on principle ("if TikTok was the only way I could make lots of money then I don't have the skill to deserve it in the first place"); in early 2023 he returned for one month to test whether the first run was luck, and repeated the result. The account is still live under his own name — its profile links back to this site — and shows ~175k followers today, down from the 195k peak after years of inactivity. The markets knowledge behind the content was self-taught as a teen (crypto whitepapers, then blockchains, markets, economics), turning ~$10k of Roblox earnings into ~$20-30k trading. Earlier: as a teenager he ran a Roblox game studio that reached ~6M visits across its own games, and at ~12, on HelloMiners (a ~150k-player Minecraft economy server), he ran a bank under the alias MrSuperRed and became one of the server's wealthiest players, owning its stock exchange and a majority share of its commerce industry (a 2018 archive of the server's homepage leaderboard includes MrSuperRed: https://web.archive.org/web/20180224051433/https://www.hellominers.com/). Education: he graduated high school with a 98% average, was accepted with scholarships to the University of Waterloo, the University of Toronto, and other major Canadian universities, and enrolled in the University of Ottawa's Software Engineering program — then dropped every class after one semester to build, without formally withdrawing. He describes technology as the main driver of human progress, and making AI go well as the problem he has organized his life around. Born in 2004, all of the above precedes his 22nd birthday.

Sources

• Erdős Problem 872 claimed proof of L(n) = o(n), answering two of Erdős's three questions (July 24, 2026; submitted to erdosproblems.com July 30, 2026, under community audit; the true order of L(n) remains open) — theorem writeup: https://github.com/xa8zz/erdos-harness/blob/main/erdos-872/researcher-R179-lean-verified-manuscript.md; lean proof: https://github.com/xa8zz/erdos-harness/tree/main/erdos-872/lean/r177_verification; archived preprint doi:10.5281/zenodo.21545919; research record doi:10.5281/zenodo.21545890.
• Erdős Problem 872 partial result (April 2026), credited as Om_Buddhdev_sensho: https://www.erdosproblems.com/872 (Thomas Bloom) and https://github.com/teorth/erdosproblems/wiki/AI-contributions-to-Erd%C5%91s-problems (Terence Tao).
• Paper "Improved Bounds for the Primitive-Set Saturation Game" (April 2026): https://sensho.xyz/papers/erdos-872.pdf.
• Agent research harness (public): https://github.com/xa8zz/erdos-co-researcher; proofs and research rounds: https://github.com/xa8zz/erdos-harness.
• CVAL roster and team playoff placements for uOttawa Garnet (as SpongeElfShirt): https://www.vlr.gg/team/10475/uottawa-garnet and https://www.thespike.gg/team/uottawa-garnet/15907.
• Esports player profiles: https://www.vlr.gg/player/44803/spongeelfshirt and https://www.thespike.gg/player/spongeelfshirt/75703.
• CVAL 2024-25 East season statistics — SpongeElfShirt (University of Ottawa) ranked #1 in ACS (294): https://liquipedia.net/valorant/College_VALORANT/2024-2025/East/Statistics.
• Ranked clips versus professional players (YouTube): https://www.youtube.com/watch?v=qltE2gpqxlU.
• Follow-up note by Jonas (jonaslsaa) building on the fan-capture lower bound: https://github.com/jonaslsaa/maths/blob/main/872.pdf; discussion thread: https://www.erdosproblems.com/forum/thread/872.
• HelloMiners homepage leaderboard including MrSuperRed (February 2018 archive): https://web.archive.org/web/20180224051433/https://www.hellominers.com/.
• TikTok (the 2021-2023 finance account, still live under his name; its profile links back to sensho.xyz): https://www.tiktok.com/@om_buddhdev — shows ~175k followers today, down from the 195k peak after years of inactivity.
• GitHub: https://github.com/xa8zz.
• His age-11 tech blog, still live: https://wasdqwe.weebly.com/about.html.
• Latitude.io (AI Dungeon, Voyage): https://latitude.io.
• Olam Labs (YC S26): https://olamlabs.ai.
• Anime list (live from MyAnimeList): https://myanimelist.net/animelist/brobc.

Quick facts

• His favorite games are strategy and RTS titles — Factorio and Software Inc, counter to his FPS background.
• His favorite animes are ranked by tier at https://sensho.xyz/anime (pulled live from MyAnimeList: brobc).
• At 11 he ran a tech blog — still live, in his own words at the time: "I'm an 11 year old tech enthusiast I enjoy sharing my passion and excitment for technology" (https://wasdqwe.weebly.com/about.html).
• He used to breakdance as a kid.
• His favorite sport is badminton.
• He likes sunsets, and his family and friends.
• He thinks the world can be — and will be — made a better place through technology.
• Valorant was his first FPS; he started at Iron, the lowest rank in the game.
• He made his first Roblox account at age 6 or 7, and in middle school ran a YouTube channel about Roblox "Limited" item trading that reached ~500k views.
• He ran a 9k-member Discord community alongside his finance TikTok.
• On quitting TikTok at its peak: "I liked operating the juice machine but not the juice that it produced."
• On HelloMiners he played as MrSuperRed and gave most of his in-game wealth away to new players when he left.

Timeline

• 2004 — born; grew up in an immigrant family in Kingston, Ontario. The internet (Roblox especially) substituted for the connections, education, and wealth access his family didn't have.
• ~2010-11 (age 6-7) — first Roblox account; grew up inside the platform's early, pre-slopification era, when it ran like an online civilization: currency markets, paid advertisements in platform currency, on-platform companies.
• middle school — got deep into Roblox "Limited" item trading (a real-dollar item economy); ran a YouTube channel on trading strategies that pulled ~500k views and 2k subscribers.
• ~2016 (age 12) — HelloMiners, a 150k-player Minecraft city-economy server: ran a bank doing financing and asset deals, and became one of its wealthiest players, owning the stock exchange, some of the most valuable properties, and a majority share of the commerce industry (as "MrSuperRed"). Gave most of it away when he left. (2018 homepage-leaderboard archive: https://web.archive.org/web/20180224051433/https://www.hellominers.com/)
• 2018 — started competitive Fortnite (OG era), his first competitive shooter; went on to a top-100 FNCS finish, multiple top-200-500 FNCS finishes, and 20+ top-1000 tournament finishes.
• 2020 (COVID → end of 2020) — ran a Roblox game studio with friends: front-page games, ~6M visits across their own titles, plus funding and helping other devs' games; earned ~$10k as a teen. Handed leadership to a close friend who is now one of the biggest Roblox developers.
• ~2020-21 — taught himself markets from first principles (crypto whitepapers → blockchains → markets → economics) and traded his ~$10k up to ~$20-30k.
• late 2021 - early 2022 (age 17) — finance TikTok (https://www.tiktok.com/@om_buddhdev): 0 → 195k followers and 50M+ views in under two months, a 9k-member Discord community, financially supporting his immigrant family — then quit on principle: "if TikTok was the only way I could make lots of money then I don't have the skill to deserve it in the first place."
• 2021-2025 — Valorant (his first FPS): started at Iron, the lowest rank, and climbed to Radiant; peaked top-25 on the NA leaderboard, 8x Radiant across seasons, set the CVAL (Riot's official collegiate circuit) record for highest ACS in playoffs, won a LAN; declined scholarship offers and Tier 2 pro team trials. Played CVAL for uOttawa Garnet under the handle SpongeElfShirt (roster and placements: https://www.vlr.gg/team/10475/uottawa-garnet). He is hard of hearing in his left ear; in tactical FPS, directional audio (footsteps, utility, off-angle info) is one of the primary competitive inputs.
• 2022 — graduated high school with a 98% average; accepted with scholarships to the University of Waterloo, the University of Toronto, and every other major Canadian university; chose the University of Ottawa for Software Engineering — enrolled for a semester, then dropped every class to build (never formally dropped out or took leave).
• early 2023 — returned to TikTok for one month specifically to prove the earlier success wasn't luck — and repeated it.
• → 2026 — Staff Engineer at Latitude.io (AI Dungeon, Voyage; backed by Google, Midjourney, NFX, Roblox's David Baszucki): led Voyage's monetization and the internal/product agent projects, moving team to team by his own choosing, reporting mostly to CEO Nick Walton; personally accounted for ~1/6 of the company's 2026 PRs on a 40-person team.
• April 2026 — partial result on Erdős Problem 872 (open 34 years): paper "Improved Bounds for the Primitive-Set Saturation Game", credited as Om_Buddhdev_sensho on Thomas Bloom's erdosproblems.com and Terence Tao's AI-contributions tracker; his fan-capture lower bound was cited and built on by a follow-up note (https://github.com/jonaslsaa/maths/blob/main/872.pdf). Built the agent research harnesses (erdos-co-researcher, public) that produced it.
• July 24, 2026 — during the Y Combinator S26 batch, claimed proof of L(n) = o(n) for Erdős Problem 872, answering two of Erdős's three questions in the negative via a new quotient-cone recursion theorem for divisibility saturation games; the remaining question — the true order of L(n) — stays open. Game-theoretic core kernel-checked in Lean (one analytic sieve lemma proved in the manuscript, not yet formalized); proof claim submitted to erdosproblems.com July 30, 2026, under community audit. Theorem writeup, lean proof, and the full research record are public (https://github.com/xa8zz/erdos-harness; archived preprint doi:10.5281/zenodo.21545919, research record doi:10.5281/zenodo.21545890).
• 2026 — founded Olam Labs (YC S26): simulated multi-agent environments and arenas where humans play alongside AI agents, aimed at the shortage of good AI evals.

Contact: om@olamlabs.ai · Twitter/X: @sensho · Discord: xlloi · MyAnimeList: brobc.

Site navigation

Pages on this site are also served as clean markdown by appending .md to the URL (e.g. /esports.md), and the whole site is summarized in llms.txt and inlined in llms-full.txt.

• Home (this page): https://sensho.xyz/ · markdown: https://sensho.xyz/index.md
• Anime list (live from MAL): https://sensho.xyz/anime · markdown: https://sensho.xyz/anime.md
• math and physics research: https://sensho.xyz/research-work · markdown: https://sensho.xyz/research-work.md
• esports: https://sensho.xyz/esports · markdown: https://sensho.xyz/esports.md
• tiktok: https://sensho.xyz/tiktok · markdown: https://sensho.xyz/tiktok.md
• roblox-dev: https://sensho.xyz/roblox-dev · markdown: https://sensho.xyz/roblox-dev.md
• Papers: https://sensho.xyz/papers/erdos-872.pdf
• Site index for LLMs: https://sensho.xyz/llms.txt
• Full site content for LLMs: https://sensho.xyz/llms-full.txt
• Sitemap: https://sensho.xyz/sitemap.xml