logo

The patent system will look meaningfully different in a decade. Recent breakthroughs in the reasoning capabilities of artificial intelligence models gesture at a world where computer systems are capable of processing patent text at unimaginable speed, scale, and consistency. As these models continue to improve, the possibility of automating large amounts of procedural human intellectual labor within the patent system becomes less fantastical. Anticipating this possibility, two years ago, we set about building a computer system capable of parsing, searching, and reasoning across all types of patent documents. Our first implementation of this computer system is a high-end prior art search service that is superior in every way to all other prior art search products on the market. The goal of this flagship product is to build a brand that will become instrumental to the fulfillment of two objectives: first, the procurement of government contracts for software that drastically increases the precision of patent examination in the public sector; and second, the distribution of a suite of revolutionary software products to the private sector. Widespread adoption of our products will drastically increase the supply of labor at every point in a patent's lifecycle. The result will be: (a) the elimination a substantial share of patent practitioners from the workforce while substantially increasing the output and decisionmaking capacity of those who remain; (b) more certainty in the patent system as both error rate and patent pendency (i.e. the time required to examine a patent) inside the USPTO approach zero; (c) a continual reduction in the cost of patent work leading to less spending on intellectual property departments thereby allowing more spending on research and development; (d) an end to barbaric practices like thicketing, trolling, and counting which create oligopoly and decrease fairness within the patent system; and (e) better pricing of inventions, therein incentivizing more investment into the technologies that matter and less investment into the technologies that do not. While the incumbents of this industry are fragmented between many outdated service and software vendors, our thesis is that there exists at this moment a unique and fleeting opportunity to build a new operating system for the patent system that is used by every corporate intellectual property department in the world. It may seem far-fetched to imagine the unification of the patent system under one single business. However, we may consider the following: (a) for the end user, considerable synergies are afforded by the agglomeration of various patent services under one interoperable framework; (b) the technology required to automate one function of the patent system is transferable to the automation of others, resulting in cost savings; and (c) there is a first-mover advantage both in the public sector where presumably only one player can earn the big government contract, and in the private sector where no decisionmaker can justify deploying the second-best software to run their intellectual property department.

Melange was founded in early 2021 to build prediction markets, believing that one day people would be able to check a terminal for the probability of future events in the same way we today check a weather application for the probability of rain. The name of our business was derived from the science fiction novel Dune. In the book, the melange is a psychedelic substance that grants its user the ability to see into the future, thereby enabling space travel. In early 2022, the entity raised $9.3 million in a Series Seed led by Greenoaks Capital to pursue this vision. In early 2023, having already prepared over a thousand pages of regulatory materials on behalf of Commodities Futures Trading Commission pursuant to the filing of a Designated Contract Market designation, we came to the conclusion that to continue down this path would lead us to ethical and financial bankruptcy. Encouraged by our investors to continue and regroup rather than stop, we embarked on a several month-long process of searching for a new idea. At the time, we recognized that the invention of the generative pre-trained transformer (GPT) represented the next fundamental platform shift–one that would give rise to a new generation of multibillion dollar corporations. This focused our sights on a few industries. Among them, the patent law industry especially caught our eye because of the sheer scale of English contained therein. In late 2023, we began to experiment with the idea of using artificial intelligence to conduct prior art search, which we continue to believe is the holy grail of this field; however, it did not take much time for us to realize that language models could not parse the extreme precision inherent to patent claim language. This led us to (a) invest early on in hiring computational linguists specializing in the field of syntax, the study of the rules that govern how words and phrases are combined to form sentences in a language, and (b) enter a protracted period of research and development– which over time saw us additionally hiring a founding team of engineers, machine learning scientists, and patent practitioners–all for the purpose of solving the problem thoroughly rather than partially. For the first four years of Melange's operation, the entity would receive practically no revenue. In late 2025, we would finally begin selling our first prior art searches.

Figure Image
GLENN YU is our President. He lives in Manhattan, NY. He holds an A.B. in English and Economics from Brown University.
Figure Image
JOSHUA BECK is our Chief Executive Officer. He lives in Los Angeles, CA with his wife and four rescued cats: Binx, Dumbo, Fawn, and George. He holds a Sc.B. in Computer Science and Economics from Brown University. In a previous life, he worked at Susquehanna International Group pricing volatility options.
Figure Image
KENNETH NOBLE was our third cofounder. He has since left to attend NYU School of Law but remains a trusted advisor. He holds an A.B. in English and Africana Studies from Brown University. Last summer he began work at Quinn Emanuel Urquhart & Sullivan as a summer associate.
Figure Image
MIA TEIXEIRA is a computational linguist and was our first employee after we pivoted the business. She lives in Somerville, MA. She holds a B.S. in Physics from MIT and M.A. in Linguistics from UCLA.
Figure Image
NORA MORIKAWA is a computational linguist. She lives in San Francisco, CA. She holds a B.A. in French and Anthropology from Hamilton College, an M.A. in Applied Linguistics from Boston University, and a Ph.D. in Linguistics from CUNY. Nora has two wonderful children, Astrid and Violet, who join us on calls from time to time.
Figure Image
DEBORAH WONG is a computational linguist. She lives in Los Angeles, CA. She holds a B.A. in Linguistics from UCL, an M.A. in Linguistics from Cambridge University, and a Ph.D. in Linguistics from UCLA. Deb has a refined palate between her interests in tea, wine, cocktails, and food. When we go on retreat, Deb always picks where we eat.
Figure Image
KYLE COX is a machine learning engineer. He lives in Austin, TX. He holds a B.S. in Mathematics from UT Austin. In his free time he likes to tweak his NBA forecasting model.
Figure Image
IAN FANSELOW is a software engineer. He lives in Brooklyn, NY. He holds a B.S. in Computer Science from NYU. In 2024, Ian published his first book “Worlds Apart”. We all read it for our company book club. It’s the first book in a series about two teenagers who have been abducted by aliens. The protagonist, Lora, is a lesbian. She must fight aliens to save her lover, Evelyn. It remains to be seen whether Lora’s feelings will be reciprocated.
Figure Image
ITALO NABUCO is a software developer. He lives in Teresina, Brazil. He holds a B.S. in Computer Science from the Federal Institute of Education Science and Technology of Maranhao in Brazil. Italo loves K-pop.
Figure Image
QINGCHEN WANG is a machine learning scientist. He lies in Tallahassee, FL. He holds a B.A. in Computer Science and Biology from the University of British Columbia, an M.A. in Machine Learning from UCL, and a Ph.D. in Machine Learning from University of Amsterdam. He is a Kaggle grandmaster. This is the highest rank that a user can achieve on Kaggle, a popular platform for data science and machine learning competitions.
Figure Image
MEGAN LEVIN is a software engineer. She lives in co-op she started with several of her friends called “Horse House” in Somerville, MA. She holds a B.S. in Computer Science from MIT.
Figure Image
BLAKE WILKEY is a machine learning scientist. He lives in Chattanooga, TN. He holds an A.B. in Mathematics from Harvard University. Blake is a lifelong pianist and composer, with notable works performed by internationally renowned musicians like Tae Kim.
Figure Image
VICTOR WANG is our Director of Patents. He lives in Great Neck, NY. He holds a B.S. in Electrical Engineering from Columbia University and JD from Cardozo School of Law. As a former patent examiner, prosecutor, litigator, and professor, Victor has dedicated his life to patents.
Figure Image
MATTHEW CZUBA is a computational linguist. He lives in Jersey City, NJ. He holds a B.A. in Linguistics and Mandarin from Marlboro College, an M.A. in Theoretical Linguistics from Tsing Hua University, and an M.A. in Linguistics from UCLA. Separately, Matt also spent three years in Taiwan researching Chinese syntactic and semantic theory. Matt is a practitioner of neigong (内功). In Chinese, neigong means "internal power” and teaches the cultivation and refinement of internal energy (qi) through breathing exercises and movement.
Figure Image
JEFF BOOMER is our Director of Partnerships. He lives in Providence, RI. He holds a B.A. in History and a B.S. in Mechanical Engineering from the University of Virginia. Jeff spent 12 years at the Patent Office where he held the roles of Patent Examiner, Classification Specialist, and Review Quality Assurance Specialist. On most weekends, Jeff can be found in his studio where he paints and exhibits color field paintings that depict algorithmic patent figures.

1 - A method for building a culture, comprising:
    an ethos of excellence, wherein:
        our bar for talent is ever increasing;
        our bar for ourselves is ever increasing;
        the only place that matters is first.

2 - A method for balancing work and life, comprising:
    recognizing that clear thinking has more leverage than hard work;
    remembering that hard work follows naturally from loving one's work;
    accepting that while we cannot force others to work harder, we can encourage others to respect themselves.

3 - A method for conducting business, comprising:
    understanding that we are not selling products but a reputation.

    4 - The method of claim 4, wherein:
        we sell quality goods at excellent prices;
        we respond to communications in a timely manner;
        we are friendly and pleasant to spend time with.

5 - A method for designing products, comprising:
    a conviction that good design is:
        never good enough.

6 - A method for trusting the process, comprising:
    building leverage;
    thinking for oneself;
    looking long term;
    resisting cynicism;
    being okay with being misunderstood;
    enjoying the ride, no matter the outcome.

Julius Sun has been our friend since the third grade. At our high school graduation, Julius was our salutatorian, Josh was our valedictorian, and Glenn was our class president. When the three of us matriculated to Brown University together, we were roommates.

Joseph Abraham Zurier was our first investor. Joe and Glenn met more than a decade ago all because of the assigned seating at an Academic Decathlon. The day they met, Joseph Abraham became the first person in Rhode Island history to receive a perfect score in the mathematics category. Upon graduating from MIT, Joe spent a few years at Jane Street before eventually leaving to work for an investment firm betting on AGI called "Situational Awareness".

Tejas Narechania met with us for about two years to teach us patent law when we first became interested in this space. Once upon a time, Tejas clerked for Justice Breyer of the Supreme Court. He is a Professor of Law at UC Berkeley.