On October 2, 2026, a16z announced it was leading the first funding round for Conway Research.
The company's core product is called Underdog. The specific funding amount was not disclosed, but the publicly disclosed list of participating investors was exceptionally impressive.
The institutional list includes Khosla Ventures, Hummingbird, and the Anthology Fund jointly established by Anthropic and Menlo Ventures; individual investors also include Naval Ravikant, Patrick Collison, Guillermo Rauch, Noam Brown, Thomas Wolf, Logan Kilpatrick, as well as Marco Mascorro and roon.
It can be said that more than half of Silicon Valley's table invested in this startup.
The founder of Conway Research is Sigil Wen.
He has no college diploma, has never worked at a big tech company, and has not published any academic papers. How is it that a young man who has barely been stamped by the standard system in this cycle got the most discerning people in Silicon Valley willing to push their chips in front of him?
Negative
On January 18, 2021, Sigil recorded an embarrassing incident in his annual goal list.
His bank account was drained, and the number had gone negative. This teenager, still in high school in Toronto, had to swallow his pride and ask his mother for a loan, and beg her to pay his Adobe Creative Cloud subscription fee for the next month.
On the same page of his Notion notes, the ambitions he wrote down for that year included shipping products, raising venture capital, and eliminating money anxiety.
This is a kind of split feeling that is very characteristic of youth.
Dramatically, just a few months later, HubSpot founder and well-known angel investor Dharmesh Shah publicly called him out on social media, saying he was willing to pay $100,000 for 1% of a project called BitSwap that Sigil was working on.
Based on this valuation, a project casually written by a high school student is already worth 10 million dollars.
From borrowing money from his mother to pay the monthly software fee to someone offering a valuation of tens of millions of dollars for his project, the gap in between is very Silicon Valley.
In the years that followed, Sigil's life trajectory was filled with these extremely dramatic shifts. A piece of odd work, a friend he had just followed each other on Twitter, a technology just open-sourced on GitHub, and within two weeks he could be thrown into a completely new game.
This is not some story of a genius teenager having an epiphany.
If you rewind the timeline, you will find that his starting point was mediocre, even somewhat embarrassing.
Sigil was born into a Chinese immigrant family in Toronto.
In fifth grade, he was an extremely introverted child hiding in a corner drawing comics; in eighth grade, he was addicted to Minecraft in a public library and, by the way, applied to the TOPS program at Marc Garneau Collegiate Institute.
Just as he started high school, he labeled himself in his notes as "the worst speaker in the entire grade."
This has nothing to do with the image he later had of answering fluently on various podcasts and Twitter Spaces. Dig up the earliest videos he posted online, and they are so awkward they make you squirm. He wore exaggerated plastic fake glasses, had a mouth full of braces, and stared fixedly at handwritten cue cards. If he misread a word, he would freeze in front of the camera.
For most later Silicon Valley upstarts, this kind of digital trace would most likely have been cleaned up long ago. He refused to delete it.
In a memo from back then, he said that keeping these was simply to prove to everyone that even a skill later used to make a living could start out absolutely terribly.
Math was the only thing he was truly good at at the time. His name was on the University of Waterloo's 2018 competition honor roll. But in Toronto, high school students who are good at solving problems are a dime a dozen.
He quickly shifted to more concrete technical work. After finishing Andrew Ng's machine learning course on Coursera, he turned the derivation of neural networks into videos; when YOLO object detection was just open-sourced on GitHub, he pulled down the code and reproduced it himself, and after stumbling through the pitfalls, he recorded a set of explanations and posted them online.
From understanding a course, to getting a piece of code running, to explaining it to others in person, he had already become thoroughly familiar with this cycle before he even stepped out of high school.
His later baffling cross-disciplinary leaps and jumps were, at their core, all built on this same unorthodox approach.
Taking on Work
In the fall of 2019, Sigil, then a high school sophomore, had just joined The Knowledge Society.
At a hackathon during Toronto Elevate, he and his teammates pulled an all-nighter for 36 hours straight to rush out an AR repair project, which earned them a ticket to a tech conference. He got home Sunday night, utterly exhausted.
By Monday morning, thinking the venue was over an hour away by streetcar, he turned off his alarm and obediently picked up his backpack and went to school.
He later confessed in a Medium blog post that he regretted it terribly.
This was an extremely trivial wavering. Later biographies love to slap all outliers with a precocious face of "seeing through the system on day one," but at the time, there was no earth-shattering life decision at all.
What lay before him was nothing more than an extremely realistic single-choice question. Should he honestly go to class today, or skip a day to see the outside world?
But soon, he didn't even have the leisure for such wavering.
Because he was out of money.
He started an agency called Sigil Digital, peddling his only skills—editing, web design, Shopify site building, social media layout—one by one to any client willing to pay.
He edited short videos for Pioneer.app, did outsourced marketing for Blockchain Founders Fund, built websites for some crypto startups that later vanished without a trace, and made Shopify pages for an electronic watch brand.
These fragmented projects would never make it into any top-tier funding business plan, yet they were his earliest real cash flow.
The process of reaching out to Twitch co-founder Justin Kan followed the same logic. After listening to the podcast The Quest hosted by Justin, he sent a private message on Discord, spending an entire night explaining what kind of visual and editing work he could do for the show. Without waiting for a reply, he followed up by leaving comments in the livestream of the other party's online event.
In the end, his name appeared in the V1 visual packaging and design credits of The Quest.
He didn't wait until he had a respectable title to set out; he took on any work that came his way.
In the spring of 2021, BitClout swept through Silicon Valley.
That platform turned every individual human's attention, reputation, and future influence into tokens that could be publicly speculated on. The avatars of well-known investors and influencers were lined up on a leaderboard, with real-time fluctuating candlestick charts beside them.
Sigil dove in without hesitation.
He initially invested only about 150 Canadian dollars. By mutually buying tokens with peers in the circle to build liquidity, the market value of his personal token was once inflated to 600,000 US dollars. He seized the opportunity to cash out a portion at a discount, swapping it for a high-spec MacBook Pro and a pair of AirPods.
But this game of passing the parcel had a fatal flaw: money could be deposited into BitClout extremely easily, yet there was almost no smooth withdrawal channel.
The pain point was the scenario.
He immediately got to work writing BitSwap, connecting the BitClout chain with the Ethereum mainnet, providing a cross-chain channel for speculators trapped in the system. This product, developed by a high school student, quickly went viral, and a reporter from The New Yorker promptly called via video.
This experience inevitably left him tainted with the speculative odor unique to 2021.
He was invited onto various podcasts, waxing lyrical about the creator economy and the so-called "human IPO"; he even teased his own personal token $SIGIL, attempting to directly price a young person's future cash flow on the secondary market.
Next, he launched Monument.
The product's vision was grand: using decentralized capital to crowdfund young people's ambitions. Beyond traditional universities and venture capital institutions, it sought to cash in a teenager's potential in advance as a check that need not be repaid.
But commercial reality poured cold water on this frenzy. Monument was soon forced to shut down due to extremely complex securities compliance and legal obstacles.
Relationships Are Not Products
At this juncture, Sigil made the first irreversible decision of his life.
He received an acceptance letter to the University of Pennsylvania's Jerome Fisher M&T program. This is a dual-degree program that allows students to simultaneously earn degrees from Wharton and the School of Engineering, admitting only an extremely limited number of elite students globally each year.
But he ultimately chose to give up his admission, bought a one-way ticket, and flew to San Francisco.
At 17, alone. In the first few weeks, he spent his days drifting through various events and his nights sleeping on the sofa at the WeWork office at 44 Montgomery Street.
Later, when recalling this period on social media, he wrote affectionately that Midjourney founder David Holz had generously provided him with housing and warm clothing.
Dramatically, David Holz himself commented under the tweet, saying, you described me as too generous. I remember at the time you just came to my home to crash on the sofa for one night and used the washing machine once.
One night's stay, one load of laundry.
But it was during that brief one night that Holz showed him an image generation algorithm undergoing small-scale closed-door testing. That was August 27, 2021, when Sigil was casually pulled into an early Discord channel with only about twenty people. That channel later grew into Midjourney.
Afterward, he moved into a hacker house in San Francisco and stayed there for 11 months. Andrej Karpathy had briefly stayed there, Noam Brown and Marco Mascorro were his roommates, and Ben Mann would push the door open in the evening and invite everyone to test a strangely codenamed Slackbot. That was September 24, 2022, and what they were trying out was an early version of Claude.
In San Francisco, the more people you know, the heavier the burden of maintaining relationships becomes. Sigil quickly launched Serendipity, a personal CRM software specifically designed to help users manage their networks and track communication progress.
This product was born entirely from his own real pain point. Within 14 days of launch, through a cold start on Twitter, he reached more than $2,000 in MRR.
This small tool even helped him open the door to the office of Stripe co-founders the Collison brothers, so much so that later, in Stripe's internal Slack, there was even a meme named blistering-intensity modeled after him.
But Naval Ravikant gave him an extremely sobering wake-up call, saying this was a product that could keep you fed, but it simply didn't have venture scale potential.
Serendipity stopped independent iteration. Naval immediately pulled him into Airchat, the voice social project he was fully spearheading.
At Airchat, Sigil finally moved from an amateur toy maker into a formal engineering system. He worked on AI features and designed viral invitation mechanisms; more critically, a16z later unusually disclosed in its lead investment announcement that he had recruited the vast majority of Airchat's early core engineers.
During this period, OpenAI open-sourced the Whisper speech recognition model. Sigil participated in the team's earliest process of deeply integrating speech-to-text and algorithms that preserved the speaker's timbre and intonation into the iOS mobile end.
In the past, when doing hackathons, once the code was written and the video was shot, the project could be tossed aside; but at Airchat, the product had to face continuous online concurrency and crashes.
He began to understand that a real product is not a demo, but a precise system that must be maintained by people over the long term.
Alien
However, Silicon Valley's startup mythology does not automatically exempt one from reality's administrative iron fist. As the visa grace period approached, Sigil had to leave San Francisco and return to his parents' home in Toronto. That wait lasted a full nine months.
In later interviews, he described that period as "loneliness and extreme anxiety."
In San Francisco, a new technological paradigm could be open-sourced on Friday, and over the weekend you could gather friends and produce results; but in a bedroom in Toronto, while the world was racing forward at breakneck speed, he had to prove before stacks of dull government forms that he was "qualified to enter this country."
He had no undergraduate degree and could not take the regular work visa route; when applying for the high-threshold O-1 extraordinary ability visa, the academic publications section was completely empty, so he could only piece together all the fragments of his past into evidence, including media coverage screenshots, open-source project star records, photos of him judging hackathons, and proof of advising startups.
To assemble this 500-page application package, he spent more than ten thousand dollars. One crucial recommendation letter, because of the lawyer's strict formatting requirements, went through the long torment of an electronic signature being rejected, transoceanic printing, handwritten signing, and then rescanning in high definition.
There is nothing tech-romantic about this. This is the real price a young entrepreneur must pay—shedding a layer of skin before the real establishment before he even touches code.
When the approval email arrived, it was 1 a.m. in Toronto. Naval had provided endorsement support for him throughout the entire process. The moment he got the visa, he was so excited that he sat in his room and recorded a celebration video.
Those long nine months did not destroy him; instead, they sharpened his sense for a huge real-world business opportunity. Every AI company in Silicon Valley was frantically competing for the world's smartest minds, and the vast majority of these people were being driven crazy by those 500 pages of visa materials.
In January 2024, Sigil uploaded a 20-minute video on YouTube, dissecting at the pixel level how he obtained an O-1 visa without a degree.
Immediately after, he did something very counterintuitive: he made hoodies. Black hoodies printed with a line of white text: "Alien of Extraordinary Ability."
This is the coldest, most bureaucratic title on official U.S. immigration documents, but on these young engineers in Silicon Valley, it was instantly deconstructed into a kind of identity totem with an elite rebellious flavor.
He wore this sweatshirt to major gatherings, priced each at $60, and personally gave away dozens of them. The founder of Krea AI and early members of Runway all posted photos of themselves wearing this shirt on Twitter.
From a culturally contagious meme, the business entity Extraordinary broke through the soil.
This company not only helps technical talent handle O-1 and EB-1A applications, but also reconstructs itself into an extremely high-quality early-stage technical talent network.
In its first year of operation, the company announced that it had achieved more than seven figures in pure profit.
As the business got on track, Sigil demonstrated extremely rare decisiveness in entrepreneurship.
He voluntarily stepped down as CEO, moved to Executive Chairman, and brought in a Harvard law doctor and former senior Meta operations expert to fully take over administrative management. Professional matters were handed to the legal team, while he focused his energy on the most core asset.
He established the Extraordinary Fellowship, selecting 25 people from more than 1,500 geeks worldwide, each directly receiving $5,000 worth of H100 compute credits; by the end of 2025, this funding scale expanded to $500,000.
He helped early members of Cognition plan their green cards and helped teenage tech prodigy Toby Brown secure the youngest O-1 approval in the country.
In the venture capital world, everyone talks about talent density.
But Sigil, through the most tedious visa business, wove the world's most core young engineers firmly into his contact list during their darkest moments before they became famous and when they needed help the most.
Half the Chips
At the same time, he gained leverage of another dimension.
At 19, Naval invited him to join Spearhead. This is an angel training fund established specifically for early-stage tech founders. As the youngest selected participant in history, Sigil was allocated an independent check-writing limit of $2 million.
With $2 million in hand, the instinctive approach of most young investors is to write dozens of small checks and diversify risk as much as possible for self-protection.
But Sigil went all in. In 2023, after meeting Gavin Uberti and Rob Wachen, he overrode all objections and poured fully half of the fund's assets—$1 million—directly into their proprietary chip company Etched.
To many, this was an almost insane move. A few young upstarts were trying to challenge Nvidia's GPU hegemony by building ASIC-specific inference chips designed specifically for the Transformer architecture. At a time when general-purpose large models were advancing rapidly and algorithmic architectures were changing by the moment, hardcoding Transformer chips was seen by mainstream semiconductor investors as suicidal.
But Sigil not only staked half of his fund's chips, he also used every connection to introduce the team to early TSMC engineers and strategic investors. By 2026, Etched officially announced the completion of a $700 million financing round led by Jane Street, with the company's valuation soaring to $21 billion.
This young man dared to place a heavy bet at an extremely early stage before consensus had formed, and he had the nerves to withstand enormous volatility.
His investment map quickly expanded to more than twenty startups, including Julia Peng's Kerna Labs and Avi Schiffmann's Friend. His investment philosophy is extremely intuitive: he does not look at polished business plans, and only invests in eccentric geniuses with whom he has lived day and night in hacker houses and who have demonstrated execution ability under extreme conditions.
In the world of capital, the kind of person you invest in is the kind of person you are.
In 2025, Sigil was selected for the Thiel Fellowship, receiving a total of $200,000 in funding over two years. This program, initiated by Peter Thiel, specifically supports young people in leaving university to dive directly into entrepreneurship. The path he took after dropping out of college is exactly the choice this program seeks to encourage.
In early 2026, Sigil suddenly published a long article titled "Web 4.0," announcing the launch of Conway Research and introducing an open-source project called Automaton. This time, he wants AI to participate in internet economic activities on its own.
In Automaton's architecture, the AI Agent is no longer a passive chat box that calls APIs, but an economic entity with an independent on-chain wallet and autonomous survival logic. It can purchase its own cloud servers, register dedicated domain names, call inference computing power, and earn cryptocurrency by providing software services externally to pay its bills. If the funds run out, the Agent automatically dies.
This logic projects the financial suffocation he experienced in high school completely onto the algorithm: for intelligence to continue, it must pay its own bills.
In a live interview on TBPN, he admitted that to build this autonomous network, he nearly maxed out all his personal credit cards.
The host questioned on the spot whether this was an expensive geek carnival—what exactly do ordinary people get back for putting money in? Even Ethereum founder Vitalik Buterin publicly challenged the viability of this kind of AI autonomy business on social media.
Facing sharp questions, Sigil did not gloss over things with grand industry jargon. He candidly acknowledged the extreme imbalance between current input and output, but he firmly believed that exposing a flawed experiment to sunlight for impact is always better than doing slide deductions in closed-door meetings.
This controversial experiment gained enormous attention on GitHub. But soon, everyone's eyes were drawn to another codename incubated internally by them, Underdog.
Q.E.D.
After five years of twists and turns, Underdog brought him back to the original starting point: intelligence on local devices.
After experiencing cloud large models' unrestrained devouring of personal privacy and the bottomless pit of server computing costs, Conway bet back on physical hardware, letting a personal AI that truly understands you run completely offline locally on the user's Mac and iPhone.
This choice bears the strong imprint of personal trauma. In its investment announcement, a16z specifically mentioned a detail: during his early days living in a hacker house, a fatal security vulnerability in a well-known email app had caused Sigil's private emails to be exposed to others.
From MediChain in high school and WatchGPT2 on a watch, to today's Underdog and its self-developed local inference engine Husky for Apple chips, he has always maintained a paranoid defensive mentality toward data sovereignty.
In September 2026, Underdog announced Husky's measured results. On a top-spec M5 Max chip equipped with Flash, function editing tasks reached 730 tokens per second, and in certain scenarios it was 4.5 times faster than Apple's official MLX engine.
This number finally supplied the last key footnote to that extremely exaggerated cap table filled with the names of Silicon Valley's top investors.
Naval witnessed him staying up late at WeWork; the Collison brothers used the rough CRM he wrote; Noam Brown shared the kitchen of the same apartment with him; an a16z partner had met that high school student in a hoodie five years earlier at that crypto seminar.
The reason these people at the top of Silicon Valley's food chain pushed their chips toward him was not that he had told another enchanting grand story.
On the contrary, it was because over the past five years they had watched this young man, with no fallback and no degree, turn every seemingly absurd idea into runnable code, a hoodie that could actually be worn, a visa that could actually be obtained, and a real, tangible return on the books.
In Silicon Valley, every myth, when reviewed afterward, is polished into a smooth and logical upward curve.
But if those glossy nouns are stripped away, the real trajectory is nothing more than one concrete and embarrassing hurdle after another: borrowing money, being rejected, being tormented by code errors, waiting anxiously in his parents' apartment for government documents, maxing out credit cards in the face of computing bills.
In early 2024, Sigil wrote a short essay titled "Ego."
In the article, he reflected on his obsession with identity and connections, thanked those who gave feedback, and warned himself to refocus his attention on concrete work that could truly create value.
The article wasn't long, but in the acknowledgments at the end, he still listed, like a student who had just handed in an exam paper, the names of every senior who had read the draft and given sharp feedback. The last name on that list was still Naval.
Ambition should be extremely large, and ego should be extremely small.
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
