After building and selling Heyzap back in 2016 I had free reign to dig into the question of “what should I really be working on?” and how to create something that would have a lasting impact on society. In part of this journey I’ve had the privilege to invest in over 150 startups from my Winter 2009 Y Combinator class to the most recent batch. While researching areas of interest, I often wondered why we have such a lack of organization of human knowledge. This gap is especially apparent for emerging technologies, new startups and ideas.
Golden’s mission is to collect, organize and express 10+ billion topics in an accessible way, presented in neutrally-written and comprehensive topic pages.
In order to help achieve this mission we raised a $5 million seed round from a16z (led by Marc Andreessen), Gigafund (Luke and Steve), Founders Fund (led by Cyan Bannister), SV Angel, Liquid 2 Ventures/Joe Montana, Aston Motes (1st employee at Dropbox), Christina Brodbeck (1st designer of Youtube), Lee Linden, Immad Akhund (my cofounder at Heyzap, now CEO of Mercury), Josh Buckley, Howie Liu (CEO of Airtable), James Smith (CEO of Bugsnag), James Tamplin (Founder of Firebase), Jack Smith, Mike Einziger of Incubus, Sumon Sadhu, Paul McKellar (Square founding team), Trip Adler (CEO of Scribd) and many other great angels.
Mapping leading edges of knowledge — from the quark to the jaguar
An abundance of knowledge is available to us via the Internet, but navigating the cacophony of formats, designs, sources and standards is challenging. Google, Wikipedia, Bing, DuckDuckGo, Quora, StackExchange, Github, etc are our go-to places for seeking information. However, the knowledge reference products in the market today fall short of mapping, in a highly organized way, everything we collectively know as a human society and what exists on the web.
Specific examples of topics that have been hard to find canonical information on include: Consensus mechanisms, new legal instruments like the Simple Agreement for Future Tokens, new fund constructions like cryptocurrency hedge funds, Arrington XRP Capital; companies/projects like Benchling, HelloSign, CryptoKitties, Ginkgo Bioworks, Momentus Space, Brave web browser, Google AutoML, Algorand; academic areas e.g., Quantum Artificial Life and Lenia, CRISPR/Cas13, phage-assisted continuous evolution, constructor theory, to name just a small sample.
It is now possible and necessary to map everything that exists systematically and pull together academic papers, documentaries, videos, podcasts, further reading, guides and meta-data that surrounds these topics.
Notability, validation and the Knowledge Age
I’m a huge fan of Wikipedia and use it multiple times per day. I believe it is one of the most important things that humans have ever built. However, as we speak, pages like SV Angel, Benchling, Lisk, Urbit and many others are and have been deleted from Wikipedia because they don’t reach the notability standard of the website.
Product Hunt was recently up for deletion but somehow survived the cull. Additionally, HelloSign, a few weeks before Dropbox acquired it for $230 million was removed from Wikipedia. BuildZoom was removed 5 times from Wikipedia and now doesn’t exist there. It doesn’t just apply to company pages though. Actual technologies, projects, products, theoretical electrical components and academic ideas are missing as well.
The arbitrary threshold of what is notable and what is not doesn’t cut it in the Knowledge Age. There are currently 5.8 million English language articles in Wikipedia, and Google had 1 billion objects (200x Wikipedia’s size) in its Knowledge Graph when it launched in 2015. We estimate internally that there are 1000x the entities to cover than what Wikipedia has today. It’s an exciting challenge!
We live in an age of extreme niches, an age when validation and completeness is more important than notability. Our encyclopedia on Golden doesn’t have limited shelf space — we eventually want to map everything that exists. Special relativity was not notable to the general public the moment Einstein released his seminal paper, but certainly was later on — could this have been the kind of topic to be removed from the world’s canon if it was discovered today?
Notability restrictions of our current encyclopedias limit the availability of knowledge around fascinating and useful niche topics. For this reason, we are taking a knowledge maximalist perspective. Let’s unleash hidden knowledge, using existence as a framework for whether a topic gets covered and supplement what Wikipedia has built over the last 18 years. Like a Mandelbrot set or squiggles from Fermat, the richness is in the edges and the margins and the footnotes.
Today, we have a great opportunity to use new technologies to solve the problems: human in the loop AI systems, fully automatic and unsupervised language writing systems, WYSIWYG editors, React, cheap cloud storage/processing, auto spelling/grammar checking, graph databases, real identity and collaboration models like Git pave the way to build a new system for compiling and serving the knowledge. With that context in mind, we want to introduce you to Golden.
Encouraging deep, extensive knowledge
Golden has been designed from the start to encourage a deep, extensive view on knowledge. If an extremely niche topic is valuable to just a handful of people and positively contributes to society, it will have a home on Golden. Some initial areas of our focus will include core topics on science, technology, and business, but also on the leading edges of research and investigation in these fields.
A modern editor
Contributing to a knowledge base shouldn’t require technical expertise or esoteric formatting knowledge. We’ve created a WYSIWYG editor to lower the barrier to entry and encourage more people to participate globally.
We have built speedy citation tools into Golden and intend to create more affordances around claim validation. You can simply highlight any claim and add a citation that is evidence to support that claim. All of these citations can then be easily viewed and examined by anyone viewing the article on Golden. We have gone a little step further by allowing ‘high resolution’ citations i.e. being able to highlight the specific part of the claim you want to back up. Over time, we want to have tighter bindings between claims and evidence in order to improve claim verification.
Finally, if you identify a problem with a particular claim or how an article is written, Golden users can “Create an issue” to call attention to the issue. If there are multiple issues on a given article, each can be dealt with and resolved separately in its own discussion thread.
Using UI and AI to save human time and get better results
As part of the thesis from day 1, we have aimed to remove all the repetitive work, bring leverage to editing, and add a little fun to the process. For example, we have built a system called “AI suggestions” which produces modular snippets of machine-learning outputs, suggested improvements that can be made to the content with simple “yes”, “no”, or “skip.”
This could be adding a crosslink to another topic page, or adding a link to an academic paper we have found around a topic. It could also be the detection and creation of an entirely new topic because it popped up inside multiple new academic papers and news publications, or it could be the addition of a detected thumbnail for the topic. Internally, we have many types of these and will release more to the general users as they become more robust. We’ve found that it’s fun to make atomic edits on mobile or on the move in a more casual way.
We have many more suggestion types coming out soon in order to increase the leverage of human help when putting the Golden knowledge base together.
We’ve also built AI tools into the topic page editing experience. Simply click on the magic wand to enter AI-assisted mode. The AI assistant will use URLs to extract relevant information to add to the topic page. This particular feature is still early in its development but clearly shows how we want to empower human editors.
The tables in Golden also take advantage of our AI extraction tools. When you paste a URL into one of our “magic cells” (i.e., a cell marked with the wand icon) in Golden, our tool will auto-extract key information from that URL, such as the title and authors of an article, abstracts, and more. This saves a lot of time!
We have also considered what makes a rich schema around a topic in terms of the metadata that surrounds it. For example, every article has a timeline to show how that entity has evolved over time. You can post a URL into our timeline AI tool, quickly get a headline and summary about the events discussed in that link, and then modify the language to make it more neutral, canonical and encyclopedic.
Our Golden AI bot is constantly working to connect knowledge across Golden and find resources that are relevant to each entity page by crawling the web. We have an extensive roadmap in this department and if you are interested in working on this problem at Golden please get in touch.
Transparency and openness
There is a strong need for this platform to be open so we are embracing Creative Commons 4.0 for the text on topic pages and are cognisant of current standards of openness. Public topic pages will be free to access and the text available on CC 4.0. If you have any feedback on this, please get in touch with us. We also have full activity logs of changes to pages that aim to bring more transparency to edits and make it easier for our team and you to fact check, monitor, and improve edits.
Golden for companies, sustainability and acceleration
Golden topic pages are completely free and open to the public. People will be able to view, edit, or create entities on Golden forever. However, our vast knowledge base and query capability means we can also provide a number of tools and services that particularly benefit companies and organizations. The features below are designed for paid users. We believe there is a good incentive alignment to have correct information on Golden that can also serve commercial-grade research requests.
Advanced query tool
Part of building the Golden knowledge base includes identifying and developing a wide range of structured and unstructured data. Paid users can use the Golden Query Tool to search across Golden’s knowledge base to find specific information in more corporate-focused queries.
We do, from time to time, open up queries to the public that we think are useful for everyone, e.g., Y Combinator Batch W19, or a list of therapeutics companies or all the VC firms, all crypto currency projects we could find whitepapers for, regenerative medicine companies, etc.
We believe this advanced query tool is extremely useful for investment funds, large consultancies and large companies, so please get in touch if you want to experience one of the best query tools out there.
If you’ve saved a particular search query that you want to monitor, you can set up alerts to be notified if any new results match your query. Users can also set alerts on particular attributes they’re monitoring on their queries or lists. For example, you can be alerted if a company has a spinout, changes CEO, enters a new market, releases a new technology, and many other cases.
Research requests and subsidizing public knowledge
Golden aims to have complete knowledge across a wide swath of companies, people, products, technologies and beyond. If you’re looking for knowledge that’s not in Golden today, please contact us! We have high-quality researchers available to build out knowledge in areas you care about.
We’ve built this into our paid plans too. Paid users who come across topic pages that don’t have as much desired information can simply click the “Request info” button. This kicks off a process with our research team and AI bots to quickly build better and more content. If users don’t get the results they expect in the Golden Query Tool, a similar button exists to start the same process to find topics that fulfill the query.
Future for Golden
Help build a knowledge base for the world
We welcome your feedback and support, whether you have found a bug, security issue, new feature request, want to help produce public content, or have a bone to pick with our policies — please do get in touch!
We encourage each person reading this sign up and contribute knowledge to Golden. In particular, we specifically welcome contributions from users of under-represented groups in order to ensure we build the most comprehensive and diverse knowledge base possible. A good way to start is to add information about your company or organization in a neutral, third person voice (no marketing language please). If you have a dataset that you’d like to share with us and open source to help make the world’s knowledge more complete, please contact us.
If you’re not ready to write something yourself, you can still sign up and contribute by reviewing the suggestions our AI systems create for human approval. Doing this helps tune our AI systems and provides valuable feedback to make Golden grow even faster.
Finally, I’d like to thank my brilliant team and contributors of Golden for working so hard to get us to a promising V0.
NOTE: Thanks to Carla Faraguna, Jed Christiansen, Rachel Phythian, and Chris Barrett for editing and helping with the post.