Jude Gomila and the Trouble With Finding Things

In 2009, Jude Gomila had an explanation for why a startup might keep its funding in the office refrigerator. Banks were having a difficult time. Better, he joked, to keep the money indoors. He and his Heyzap cofounder, Immad Akhund, had recently come from Britain to San Francisco. A visitor wanted to inspect the refrigerated fortune. The door, inconveniently, would not open.

It was a little piece of startup theatre, performed against the decidedly unfunny backdrop of the credit crunch. Heyzap had real investors and an actual product. But the joke supplies something that acquisition announcements tend to mislay: these were people making things up as they went along, with enough nerve left over to amuse themselves.

Gomila would go on to help sell that company, finance other founders and start Golden, a business with the modest housekeeping ambition of organizing human knowledge. Today he is working on Ultrametric, which aims to automate the work of starting and running companies. The refrigerator has receded from view. His appetite for an awkward problem appears to have survived.

A game should travel

Originally from London, Gomila studied engineering at Cambridge's Gonville & Caius College. His subsequent career would wander through games, advertising, investing and artificial intelligence. Engineering offers a useful way to read those apparent detours: each business involved taking a complicated system and finding a practical point of entry.

Heyzap's early entry point was a small player that websites could embed. A game could turn up wherever a reader happened to be, rather than waiting on a destination gaming site. By June 2009, the company described a catalogue of more than 12,000 games and over 20,000 websites carrying its player. Distribution was part of the product itself.

The approach had an entertaining side effect. Interviewer Andrew Warner complained that researching Heyzap kept turning into playing Heyzap. Gomila's response was that stopping him would require blocking the entire internet. The remark was cheeky, but it also described the business plan rather economically.

In that conversation, Gomila broke online growth into channels and actions: where something travels, and what makes a user send it. Sharing, inviting and challenging someone to a game were separate mechanisms. He was interested in their small contributions, their changing contexts, the places where an invitation made sense. The user was a participant with a reason to act, rather than a number to be pursued around a chart.

“Users shouldn’t even have to do the action. It should be done for them.”

Jude Gomila, discussing discovery at Heyzap

The exit with an asterisk

Heyzap changed shape. The company moved into mobile advertising and mediation, helping developers manage advertising demand. A business that had started by putting games in more places was now helping the people who made them earn revenue. Its relationship with Fyber changed too: the companies had been partners, then competitors, before becoming part of the same group.

In January 2016, RNTS Media completed the acquisition for Fyber. The familiar headline number was $45 million. The structure was more particular: $20 million upfront, with another $25 million in cash and shares tied to performance. The distinction matters. A possible acquisition payment and a founder's bank balance are different quantities, however enthusiastically a headline may introduce them.

At the time, Heyzap brought a reported 130 million monthly active users and relationships with roughly 5,000 publishers, developers and brands. Its 21 employees were expected to join Fyber. Gomila was to lead sales. The acquisition involved a working team, customers and an evolving advertising business, all the machinery that sits beneath an exit story.

Other people's companies, his own education

Alongside operating companies, Gomila became an angel investor. By his 2020 appearance on The Twenty Minute VC, his backing of more than 150 companies formed part of the introduction. The discussion ranged from early investments to the advantages younger investors might have in discovering founders before the rest of the market noticed them.

His public portfolio today includes Airtable, Carta, Gusto, Linear, Mercury, Boom and Relativity Space. The spread is instructive. Workplace software sits beside aircraft and rockets. On his investment page, he describes backing companies at pre-seed and seed, sometimes supplying the first check. This is the stage when curiosity has to do some of the work that a financial history cannot yet do.

It also creates a rather unusual classroom. A founder who invests gets to watch many other companies make decisions while continuing to make his own. Customer acquisition, hiring, distribution and timing recur, but the circumstances vary. The education is expensive in a different sense from a university degree: the exam papers are live businesses.

A blank page can be a business plan

Golden, founded in 2017 and publicly launched in 2019, grew from a specific annoyance. Gomila would investigate a niche technology, company or academic concept and find pieces scattered across the web. A video here, a paper there, a blog post somewhere else. The information existed. Getting it together was another job.

The launch announcement described a knowledge database built by artificial and human intelligence. An editor and AI tools would help people assemble information; businesses would pay for advanced queries and alerts. In other words, the broad public ambition came with a fairly concrete commercial question: what would a professional need to find, and keep finding?

The first announced funding round brought $5 million from investors including Andreessen Horowitz, Founders Fund, Gigafund and SV Angel. Akhund invested too. The old cofounder was now backing the next company. The circle also included founders and early employees from businesses such as Airtable, Dropbox, YouTube and Firebase, people familiar with turning a small useful mechanism into infrastructure.

A knowledge graph gives that ambition a particular form. Instead of treating each page as an isolated essay, it records entities and the relationships between them. A company has a founder. An investor backs a company. A technology connects to other technologies. The connections become information that can be searched and reused, rather than something a reader must reconstruct every time.

The engineering is less tidy than the diagram. Names collide. Organizations change. A piece of text can mention an entity without describing it correctly. Golden's work included extraction and disambiguation, the business of deciding what a statement refers to and where it belongs. An apparently simple box labelled with a company name can conceal a considerable amount of work.

A map finds a destination

Golden's funding grew. A $14.5 million Series A was announced in September 2020, with Marc Andreessen joining the board. A $40 million Series B followed in 2022. Across those three announced rounds, the total was $59.5 million. The money financed a demanding attempt to turn the organization of knowledge into something machines and people could build together.

On April 24, 2024, ComplyAdvantage announced that it had acquired Golden. Its interest was specific: Golden's natural-language processing, extraction and entity-disambiguation capabilities could bring more disparate information into a financial-crime intelligence system. The team of AI and large-language-model specialists would join a wider data-science operation.

There is a useful change of scale here. Mapping knowledge sounds almost too broad to touch. Determining whether two records describe the same person or business is a narrow task with consequences. A database of relationships becomes useful when a customer has a decision to make about one of those relationships.

The deal also had a long personal connection behind it. Gomila said he had known ComplyAdvantage founder Charlie Delingpole since 2005. He joined as a board observer and special advisor; a16z became a shareholder. A company built to connect information had reached its next chapter through people who were already connected.

The watch with elastic hours

There are clues to Gomila's interests outside the acquisition ledger. Interviewer Robin Zander described a fellow circus performer with interests in robotics, space and learning. Gomila's own website supplies a more recent cabinet of curiosities: interactive projects that invite visitors to inspect an idea rather than merely read a conclusion.

One is Cirwatch, a watch concept in which daylight always occupies twelve hours and night another twelve. The hours expand and contract with the seasons. Another takes apart Kandinsky's Composition VIII, letting a viewer explore shapes through different lenses. These are personal explorations, with their own purposes. They show an interest in how an explanation feels in someone's hands.

His current company, Ultrametric, returns to a more familiar set of frustrations. It describes its purpose as automating the startup, with guidance for founder processes, vendor comparisons and tools for working through company tasks with AI agents. His Delphi profile identifies him as its CEO and cofounder. Golden is now part of his history, rather than his current CEO post.

The through-line is easiest to see at the level of the user. Find a game without visiting a gaming site. Find related information without assembling a private encyclopedia. Get through a company process without rediscovering every step. The ambitions differ, but the irritation underneath them is recognizable.

Gomila once offered the advice, “Learn the rules before you break them.” It is a suitable line for an engineer whose career has kept changing categories. Start with the mechanism. Understand why it behaves that way. Then see whether the person using it really needs to do quite so much work.