How to Disrupt Google
Google search is dying. This is a weird sentence to type given Google’s unprecedented data network effects and near monopolistic position as the default search engine for most web entry points, from Chrome and Android to even Safari and iOS. And yet, I believe the decline to be real. For at least the last year or so, I’ve noticed myself appending “Reddit” to the end of most non-factual queries. In my experience, Reddit threads provide more nuanced discussion and helpful insights for complex questions, while top Google search results are increasingly listicles or junk content designed specifically to game search engine optimization (SEO).
I’m not the only one to notice this trend. Internet pioneers like Michael Seibel, Paul Graham, and Daniel Gross have all recently lamented the declining quality of Google search results in categories like product reviews, recipes, health, and code snippets. In each of these domains, more and more users are appending specific forum names like Reddit, Github, or Rotten Tomatoes to their searches in an effort to find trustworthy, non-commercial answers.
Indeed, it seems this view is quickly becoming common wisdom among the online vanguard. When author DKB wrote a particularly eloquent essay chronicling Google’s problems – which in addition to easily gameable SEO also include too many ads, misused AI that does not ask the right question, and a lack of indexing/surfacing the long tail of obscure sites – it went viral, striking a particular nerve on HackerNews, where it currently sits as the 11th most upvoted post of all time. The Internet’s early adopters are not happy with Google search.
And while it’s easy to spell out Google’s faults, it’s much harder to design a better alternative. DKB, however, also surfaced an insight which I believe will do just that. An alternative where the most effective SEO-style way to game the system is to actually produce higher quality content. His solution: “Market Rank.” MarketRank is based on the analogy of the market – a webpage with higher market value is presumed to be higher quality. This is in contrast to Google’s original famous algorithm – PageRank – which is based on the world of academic research and assumes a more heavily cited (i.e. back-linked) webpage to be higher quality.
One way to determine market value is through a webpage’s performance in knowledge marketplaces like Reddit, Twitter, Hacker News, Stack Overflow, or Quora. The currency of these marketplaces is the “upvote” or “like.” Loosely speaking, a more upvoted or liked webpage should be higher quality. And a few comparisons of Google vs. DKB’s MarketRank search results for queries like “book review zero to one” and “startup ideas” suggest that MarketRank produces more useful results – at least according to his subjective taste (and mine as well).
I believe that MarketRank is a genius, paradigm-shifting idea. But DKB is missing at least one critical piece of the puzzle: how to best normalize the value of likes and upvotes across space and time. I.e. How do you compare a like on Twitter vs. an upvote on Reddit? And how do you adjust for a site’s popularity or user inflation over time?
One effective way to solve this problem would be to measure value across sites with a universally recognized currency of the Internet. A currency which could be easily and cheaply used anywhere on the web, moving seamlessly between previously walled gardens. Fortunately, I believe that just such a currency has already emerged: bitcoin. And it can now move nearly freely and instantly on its own rapidly maturing p2p payment rails: the Lightning Network.
For the first time in history we can now attach value to information. By upvoting content with satoshis (the smallest unit of bitcoin), we can add a new dimension to a previously flat web: a “satoshi depth layer.” A web page with a higher satoshi stack is likely to be higher quality.
The next question is how best to associate satoshi depth with knowledge. One way to do this would be through existing knowledge marketplaces. Sites like Reddit or Twitter could adopt satoshi tipping for their already thriving communities. While theoretically possible, building and prioritizing an excellent satoshi tipping experience seems like too much of a foundational context switch for such a large company to make. To their credit, Twitter actually did move to enable satoshi tipping last year under former CEO and Bitcoin supporter Jack Dorsey. But since his departure, further development and promotion of this feature seems to have stalled. Reddit meanwhile is experimenting with the adjacent idea of Reddit Gold, a currency which is still trapped in its own walled garden.
A more promising approach is to build a new knowledge marketplace designed with satoshi tipping as a foundational feature from day one. By far the most impressive example I’ve seen is Stacker News, which is like Reddit or Hacker News, but where “upvotes” are measured in satoshis. Users are incentivized to post content on the site because they receive the upvoted satoshis. Other users are incentivized to upvote content with satoshis both because of altruism (which many are calling value 4 value) and economic incentives – early upvotes on content that ends up being successful yield a higher payout of rewards from the site. These rewards are currently generated from a jobs board and boosted content.
Early usage on Stacker News is very encouraging with ~1k users regularly posting and upvoting content for ~50k unique viewers. This ratio of active users to lurkers is slightly better than the well-known 1% rule for Internet forums. And given founder Keyan’s consistent shipping velocity, I expect to see much more incentive structure experimentation that should further boost site usage and signal.
One such particularly exciting experiment is the early implementation of a web of trust (WoT), whereby the value contributed by users with higher reputation scores weighs more heavily in the site rankings. Once StackerNews exposes its usage APIs, any user will be able create their own custom mashups of MarketRank and WoT. My hunch is that we’ll see a flourishing marketplace of Stacker News interfaces with different rankings based on user preference. This starts to sound a lot like Jack Dorsey’s vision for what Twitter could look like as an open protocol.
Other projects are using similar mechanisms for different media. Lightning enabled podcast players like Fountain and Breez are demonstrating early traction on the audio front by allowing users to show their appreciation or “upvote” good content with tips to creators. To date, we’ve seen >9k podcasts enable Lightning tipping and at least one Lightning enabled podcast reported that >25% of its audience is listening on value 4 value apps. Fountain even launched a “listen to earn” feature where podcasters can pay their audience to listen as well. These features not only allow for more direct p2p commerce between creators and fans – which incentivizes the creation of new knowledge – but also contribute to a very valuable, open data set of what audio content people actually value. Here’s an early example of the sort of discovery this data will enable. Sites like Wavlake are already experimenting with similar ideas for music.
Alternative approaches for assigning value to information and creating the web’s “satoshi depth layer” might look slightly or radically different. A particularly intriguing model would be to enable satoshi tipping directly on the existing web via browser extensions like Alby or Mash. Alby already enables >3k users to send satoshi tips to any Twitter or YouTube account with a lightning address in the profile. Feel free to try it out with my Twitter 😉 It truly feels like magic!
Still another approach would be to create a new search engine where users are paid a share of total site revenues to both submit queries and upvote results. The mechanism would be similar to Stacker News’, but instead of paying users to create and upvote new content, the site would instead pay users some small share of site proceeds for each search they submit (with anti-spam measures) and a larger share for upvoting results, with more satoshis paid out for early-upvoted results that end up being more popular. The next Google could literally pay you to use its service.
Despite my obvious excitement about Lightning enabled MarketRank (or ValueRank as Fountain founder Oscar Merry calls it), I do have a few major open questions:
1) How to prevent large corporations or major Bitcoin holders from gaming the system?
2) How to prevent this system from turning into a popularity contest based on who has the most followers or reach to receive tips?
3) How to build these indexes while maintaining as much user privacy as possible?
While I don’t profess to have the answers, I believe that the guiding principle for producing these answers likely lies in maximizing user choice and embracing open protocols.
In the case of preventing the very wealthy or very popular from gaming results (e.g. imagine a major tobacco company paying to upvote misinformation about the health effects of smoking or a Hollywood superstar deputizing his or her audience to support unsubstantiated medical claims), I think the best solution is to make sure that ValueRank data is open source and freely available – similar to what we discussed with Stacker News – so that a free market of algorithms can emerge, each of which uses ValueRank in a different way, likely mashed up with other data as well. One piece of likely critical data is the identity of the value contributor, ideally in a format that is decentrally issued, persistent, and pseudonymous. Such identity could be tied directly to a Lightning node or delivered via a construct like DIDs/VCs.
With these building blocks, you could imagine creating algorithms where users can choose which sites they wish to include in their ValueRank index or which users they wish to trust. David King recently proposed a somewhat similar idea when he requested a search engine that only indexes sites explicitly whitelisted by himself or people he trusts. I would personally love something like this for music, where I could discover new music based on the tips of friends whose taste in music I’ve already verified. A slightly more scalable version of this vision might include reputation staking. I could stake satoshis to vouch for someone else’s taste in music or other types of knowledge. The person I vouch for would then end up affecting the search results of people who trust me. If they end up downvoting the content recommended by this person, then I could lose my voucher stake.
When it comes to respecting user privacy while building these indexes, I think that pseudonymous identity and user choice again are the keys. While the vast majority of users will likely prefer to share their tipping data to enable a better overall search experience (this is empirically true from the status quo with Google), no one should have to participate in the creation of an index. Non-custodial web wallets and apps like Alby should be in an excellent position to enable such user choice by only collecting tipping data from those who opt in (and who perhaps are paid to do so).
All of these ideas are, of course, just the tip of the iceberg. It’s likely that the next Google will actually be built upon an insight that I have not yet considered and that the real second order consequences and challenges are of another nature entirely. What is certain though is the gravity of the current problem and opportunity. As the amount of online information grows exponentially, human attention is rapidly becoming the scarcest and most important resource in the world. I believe that using market forces enabled by micropayments with the Internet’s native currency could be the key to best protecting that attention and surfacing the right information to help drive humanity forward. The idea has justifiably captured my imagination and is now a top priority to explore and fund at Hivemind Ventures. If you are an entrepreneur or hacker exploring these ideas, please reach out!
Thanks to Oscar, DK, Ryan, Keyan, Moritz, and Kweb for feedback on this essay.
Full disclosure: StackerNews, Alby, and Breez are major positions in the Hivemind Ventures portfolio, while Fountain and Mash are personal investments made prior to starting the fund.