<center> # Don't Give Big Tech Your Papaya <big> **A strategic guide on how to avoid becoming beholden to big tech while still reaping the benefits** </big> </center> <center> *Written by Max McCrea. Originally published 2024-12-12 on the [Monadical blog](https://monadical.com/blog.html).* </center> "No des papaya," literally "don't give your papaya away," is a saying in Colombia that's typically used in the context of leaving valuables visible in a parked car, or wearing an expensive watch or jewelry while walking alone in the street at night. Big tech is getting a lot of papaya these days. For context, Silicon Valley companies love monopoly. The main reason for this is simple economics: a monopolist's share (read: all) of the economic surplus in an economic activity is a lot more than a competitors' share. You may want some of that surplus for yourself, but see that things are not moving in that direction. The ruling party in the US is now heavily influenced by a guy who wrote a book about how monopolies are inherently good and competition bad. ## Don't Depend on a Single Corporation's Services On November 30th, 2022, thousands of startups cried out in terror as OpenAI revealed that they were directly competing with applications that were building against their API. > *"I felt a great disturbance in the Force, as if millions of voices suddenly cried out in terror and were suddenly silenced. I fear something terrible has happened."* > > -Obi Wan Kenobi, AI Angel Investor This is nothing new. Amazon competes with sellers on its platform and uses data it accumulates to out-compete them[^amazon]. And even if big tech isn't competing with you, depending solely on them puts you at great risk. For example, Twitter shut 3rd party applications out of its API in 2012. [^amazon]: Ars Technica: ["Amazon reportedly used merchant data, despite telling Congress it doesn’t"](https://arstechnica.com/tech-policy/2020/04/amazon-reportedly-used-merchant-data-despite-telling-congress-it-doesnt/) Given a political landscape that's sliding rapidly towards favoring large corporations, and an age where AI might bring economies of scale everywhere, it's increasingly important to think about how to protect that papaya. ### A Rule of Two If you are building something that uses an LLM service, make sure it works with *at least two of* the major LLM providers' services. If you are building something in the cloud, make sure that it works with *at least two of* the major cloud providers' offerings. If you are working with data, make sure you are in control of it. Store and process your data using standardized formats and solutions, not proprietary ones. A happy side-effect of thinking this way is that forces you to build a better, more modular system, with redundancies in place for security & stability! ### Own Your Distribution Depending on a single channel like Google for search, or Amazon for discovery, or LinkedIn for leads, puts your business in a very vulnerable position. Instead, try to own your own distribution: * Build direct customer relationships with email lists, phone numbers, direct billing. * Develop your own customer or business community with local events, member directories, etc. * Own your platform when possible, and prefer public protocols. Self-hosted your community forums, provide direct downloads, organize on email lists, etc. * Control your monetization: enable direct payments, multiple payment processors. ### Don't Give Away Your Secret Sauce Every time you send data out to an API, you are effectively training someone else's models to compete with your business. Some examples: * Using GPT-4 for your RAG-powered copilot → Training OpenAI on your core company information and company processes * Running customer support through Google CCAI → Walking Google through the growing pains of your product or service's relationship with its customers * Processing finance docs through AWS Textract → Showing Amazon who your customers are, how much they pay, etc The key question you should ask yourself is: "If this API provider launched a competing product tomorrow, what would they learn from my usage?". Note that unstructured data, like the kind of conversation you may be having on Slack, may be the most valuable information of all as LLMs become more and more powerful[^newoil] [^newoil]: [Conversations are the New Oil](https://monadical.com/posts/conversations-are-the-new-oil.html) **Instead, look to keep proprietary workflows and domain-specific logic in-house.** Note that working with smaller firms doesn't really help, unfortunately. Smaller companies get acquired and the data they have accumulated is viewed as an asset. ### Google Doesn't Trust Anyone, So Why Should You? Google's own security philosophy is instructive here. Despite being a tech giant themselves, they treat every service - even their own - as potentially compromised: - All internal services require authentication, even between Google's own systems - Data is encrypted not just externally, but between internal services - Access is granted at the smallest scope necessary - Regular security audits of all third-party integrations This isn't paranoia - it's how one of tech's most successful companies protects itself. If Google doesn't trust Google with its data, why should you? ### A Difficult Reality Running everything in-house like Google is unrealistic for most companies. It's expensive, complex, and distracts from your core business. But the unfortunate reality is that the alternative - deep dependence on tech giants - is increasingly risky. Everyone will have to make decisions about risk/reward tradeoffs. I am certainly not advocating that everyone migrate away from using Google or Microsoft products. The middle ground? Be strategic: - Identify your crown jewels (core IP, customer relationships, critical data, internal communications), and build strong walls around these - Think carefully about information security and permissions within your organization - Accept some risk in non-core areas Just remember that as AI amplifies the advantages of scale, your papaya is only going to look sweeter.