Building Business on FOSS

Developing profitable businesses on FOSS (free and open-source software) ought to be on every VC's priority list. FOSS gives you low cost access to a global network of developers competing on the basis of merit to improve code that they care about, likely far more than any employee cares about his employer’s intellectual property.

To create and sustain an economically viable business built on code (or a protocol, or an algorithm, or a design) that is also freely available for anyone else to commercialize is a rigorous test of business acumen. But the margins and longevity that such an enterprise would obtain are ample reward for those who pass the test.


"Free" is not strictly a matter of cost

Free software does not preclude revenue. There is value to the end user in services that let him engage with otherwise "free" software in a particular manner for a fee. Also, software is not "free" merely because it can be used without paying. Code is "freer" to the extent someone cannot be stopped from using it. It is easy to see, then, that infrastructure is the critical layer to enable FOSS being that it is the physical choke point through which constraints can be enforced. In addition to being reliable (high uptime), this infrastructure should have minimal (or no) user lock-in.


Creating incentives for distributed server capacity

Imagine we are creating a Pirate Bay (anything goes) for music publishing and for listening to audio content. A "producer user" can upload any (key-signed) audio file. The user experience is like any drag-and-drop cloud storage folder, but the signing process indicates authorship of the uploaded content. Only users who access the producer’s signed file are actually consuming “the real thing,” though any user could upload the identical content with his own file signature. The file signature could also be used to encode payment instructions, for example, on the bitcoin lightning network. These payments could be sub-divided and allocated at the discretion of the author according to prescribed formulas which can be as transparent as or obfuscated as desired.

The most common "consumer users" experience could feel like any app someone would use for music discovery and streaming (e.g. Spotify). However, the code being FOSS permits a user to construct his own customized interface. The best initial user experience will likely be provided by a company with hosting resources that manages content distribution and/or lightning network services.

None of the above requires a technical break-through. The primary challenge is providing a superior user experience relative to incumbent services, so much so that a critical mass of content consumers make a behavioral switch. The cost to deploy adequate backend server resources is the economic scaling challenge. Although the early stages of development will likely require traditional venture capital funding, a globally distributed server backend could be built by allocating a portion of content payments to servers distributing the content.

Venture capital chasing this type of model using a novel crypto token probably measures in the hundreds of millions of dollars over the past five years. If, instead, a content distribution network were built with bitcoin (lightning) as its native payment architecture, the globally distributed network of media servers powering the backend of this Pirate Bay would have the same economic appeal as bitcoin mining: keep this server up and running, and you receive an average of X btc per hour.
[TO DO: calculate the break-even rate in btc/kwh of running a media server compared to an ASIC; making certain assumptions regarding the scale of servers needed per user hour, what does the break-even btc/kwh rate imply for cost in sats/minute to an end user]

User payments

  • opt-in or premium-service “sats per bit” streaming service (value streaming)
  • social media “like” (zap) with predefined payment amounts
  • “catalog” purchases
  • flexibility to allocate according to code that can be changed by the one entitled to the payment

  • it would not be enforceable as an obligation on the publisher / producer user to do so, but etiquette would favor those producers who allocate a portion of the streaming revenue to other artists whose work they've sampled

    • e.g. I sample Daft Punk to make a song of my own, and since their work is, say, about 25% of the total composition I route 25% of all payments received for this song into Daft Punk's wallet. Or, more likely, I set up a Daft Punk wallet to accrue their respective share of the bitcoin received which I turn over to them when they are ready.

Breaking the Law

Won’t people use this to break the law? Of course. But that won’t prohibit businesses and economies from emerging that are perfectly legal. Is the average person who wants to listen to a Madonna or Taylor Swift song - without paying for a streaming service or sitting through ads - going to download this FOSS Pirate Bay thing and seek out a relay that connects them to an old-fashioned torrent which happens to make the music free? Not likely.

More likely, a new Creative Commons would emerge in the uncharted waters of intellectual property law in different jurisdictions. That is, more likely than someone uploading an artist’s catalog to make available for piracy is that a content creator will use samples of copyright-protected content to create new content. This starts to get interesting. The sampling artist could choose to direct a fair and reasonable percentage of the streamed value to a segregated account for the benefit of the sampled artist (should that artist choose to claim it). But also, the sampling artist is just a signature. There need not be an associated legal identity against which to direct legal action.

Since a content distribution server can be run anywhere in the world where someone is willing to receive bitcoin in exchange for high uptime compute resources, it’s an impossible game of whack-a-mole for Taylor and Madonna to sic the police on this part of the network. As long as this piece can be properly incentivized, the business moat will endure.


Eventually, being the first to release a signed audio file in this manner will be more valuable than any legal copyright obtainable in meatspace

An obvious first business would be a music publishing service

  • key / signature management
  • content distribution network
  • promotion
    • worst case: MLM promotion and referral link schemes
    • best case: casual listeners find preferred curators to follow; the curators in turn find preferred artists to follow and promote (which could be compensated based on one of curator’s followers initiating a payment to artist)
  • premium user experience
    • you don’t publish or store any content that could be in legal jeopardy for IP infringement
    • but you can’t stop your customers from using the software to go to someone else’s servers

The AI Arms Race

A question for AI optimists. When this technology radically alters everyone's life and work as we know it, which of the following two scenarios seems more realistic to you?

In Scenario 1, you rely on an AI model that is owned by a third-party private enterprise, deployed on its servers. You have no way of knowing what might be done with your data or metadata, and you cannot audit the how the model works because this trade secret is the foundation of your vendor's astonishing enterprise valuation. In this scenario, either one group of developers working in secret maintains an R&D advantage into perpetuity or a shrewd management team manages to lock in most users (e.g. by making it inconvenient / unfeasible to switch).

In Scenario 2, you rely on an AI model that you could deploy on your own hardware at no cost because it's FOSS. You know exactly what the model is doing to transform your inputs into outputs, because you can see all of the code. In this scenario, the best models are the ones that stand up to the scrutiny of and benefit from being upgraded and maintained by the entire global network of developers working with such models.

No business with the capital to invest in frontier machine learning R&D wants to rely on another company's closed-source algorithm. Hence the largest of them are spending tens of billions of dollars on frontier R&D. How does that end? If one emerges as the clear leader, will others concede the victory and become reliant on the victor? If not, will each of the tech giants have its own proprietary algorithm? How will they all fare if a critical mass of users adopts a FOSS model (critical mass of developers => increasingly easy for users to deploy (libraries, literature) / continual feature improvements => more users => more employment opportunities for developers)?