Open Source Tools for Thin Market Automation

Deeperpoint Inc. is building Cosolvent – an open source (MIT License) AI-enabled platform to establish and automate thin markets. Our first test case is the GrainPlaza market concept using the GPSim market simulator.

Thin Markets

Thin markets, as defined by Nobel laureate Alvin Roth, are characterized by scattered buyers and sellers who have difficulty finding each other and have a low probability of random matching due to varied needs and offerings.

Why are they Thin?

Thin markets are inefficient in ways that better information exchange might greatly improve. They contrast with thick markets where many buyers and sellers exchange standardized products and services … think stock markets and Amazon.

What are Examples?

Thin markets can exist in finance, buying and selling products and services, social arrangements, and intellectual exchanges, among others. There are many thousands of real or latent thin markets. At least. Probably far more.

What are They Worth?

Globally, conventional B2B intermediaries handle $15-20 trillion annually, many in thin or thinnish markets. If automation drew even 20% more activity from these markets, that would be $3-5 trillion annually, with potential facilitator fees of $200-400 billion.

Thin Market Automation Dilemma

Thin markets face a classic chicken-and-egg problem when it comes to automation. Innovators can design a virtual marketplace, but without participants, it appears empty and unconvincing. At the same time, potential users don’t know what’s possible and won’t commit until they see the marketplace functioning. This creates a stalemate: the platform needs activity to prove its value, but it can’t generate that activity without already proving its worth.

The Cosolvent Solution

DeeperPoint’s Cosolvent is an open-source platform that lets sponsors launch new marketplaces fast. The open source portion is a hosting framework that can integrate storage and AI tools. Instead of waiting for early adopters, sponsors can then seed a simulator with synthetic users and industry context, powered by DeeperPoint’s ClientSynth service or their own methods. Custom prompts and functions make the simulator feel and behave like a live market and the client sponsor can then decide how or whether to proceed. If they go forward, genuine users will replace the synthetic ones, and the simulator will evolve into a production platform—turning a thin market into a much thicker one.

A key aspect of this design is the fact that Cosolvent does not, by itself, deliver market services. It organizes and safeguards the data while providing the utilities and connectors to the LLMs and prompts chosen by the sponsor and developer. Cosolvent is a lightweight, modular web engine for developers to build upon and it is freely available from a public Github repository under MIT License. Here’s an analogy:

If we were developing a car race, Cosolvent is the chassis, various AI LLM technologies combine to form the motor and powertrain. The Simulator is the test track. The synthetic users and industry context are the dummy passengers and cargo. The sponsor-approved AI prompts are the driver.

… Read more about Cosolvent.

GPSim – Cosolvent’s First Business Test Case

DeeperPoint’s grain market journey began with friends highlighting a specific export challenge for Western Canadian farmers. Farmers grow premium grains prized by brewers and artisan bakers worldwide, but bulk shipping too often mixes top-quality crops with lesser ones and forces sales at lower prices.

A Thin Grain Market

Western Canadian farmers grow premium grains. Direct container shipments could preserve that premium for overseas buyers, but Canadian farmers don’t know the foreign buyers, and the buyers don’t know the farms.
Chicken and egg.

Synthetic Users

DeeperPoint’s proprietary ClientSynth generates realistic market users … buyers, sellers, and logistics providers with synthesized profiles, test reports, awards, “photos” and other interesting stuff —without involving real participants.

GPSim

Our Grainplaza Simulator (GPSim), built with Cosolvent, ClientSynth and a curated library of representative real-world industry documents, creates a rich, realistic environment to test queries, operations, and add-on services.

Cosolvent/GPSim Synergy

As we develop Cosolvent and GPSim, we imagine, experiment and measure how a production version of GrainPlaza might look to the real-world market. When we find deficiencies and opportunities, we evaluate how they should be resolved. If they are a general issue or feature, we apply the lessons in the Cosolvent code. If the issues are unique to the market vertical, we build them in GPSim without revisiting Cosolvent. This creates a constant iteration between the abstract (Cosolvent) and concrete (GPSim) … round and round it goes.

DeeperPoint Business Model

From a business perspective, DeeperPoint is designing Cosolvent to be agnostic to the characteristics of any specific thin market. In our minds, it might involve money, products, services, ideas or art. With Cosolvent as a base, a prospective Sponsor (investor, entrepreneur, government agency, association, market heavyweight, etc.) can decide whether they want to explore their specific thin market idea. Then there are many ways that DeeperPoint can engage with the would-be thin market solution provider:

  • Grab and Go – A developer can clone the Cosolvent Github repo and proceed on their own. The developer needn’t inform DeeperPoint.
  • Chat and Collaborate – A developer can reach out to chat or, better still, offer to collaborate on open-source Cosolvent development. That could be totally informal, or it might involve funding and executing specific design projects or tasks of mutual interest. Call and we will brainstorm about it.
  • Hire DeeperPoint to Build a Vertical Simulator – A developer can ask Deeperpoint to help it build a simulator for their vertical as a paid service. The scope could range from providing basic technical help to building and maintaining the simulator as a turn-key system.
  • HIre Us to Help Build the Production Platform – After a developer has gathered sponsor resources and buy in, they could ask DeeperPoint to help them build the production system. Generally, DeeperPoint will be reluctant to stray so far from its main goal of Cosolvent development … but never say never.
  • Hire us as a consultant for our Generally Brilliant Advice – That’s tongue in cheek … but never say never.

Our ideal is to help as many would-be Sponsors as we can to successfully automate as many thin markets as possible.

Next Steps

  • If you are technically inclined, visit the Cosolvent Github repo and poke around.
  • Read about the details of the GPSim. We will add more detail with screenshots over time and, eventually, an interactive playground.
  • Call, Zoom or email to chat about options and possibilities.