The End Game Of OpenForage
Below is an article by OpenForage.
I’ve been really busy working on OpenForage, designing the library and all the various mechanisms behind OpenForage. It’s been really fun, but also stressful, as tall leaps of faith often are! However, there comes a time in life where a man feels like he has to stake it all on a narrow shot of success and OpenForage feels like IT to me.
It’s the culmination of everything I’ve learnt and thought about for literally decades now. It’s going to be rough around the edges as most new things often are. But I genuinely believe that in its mature form it’s going to be MORE than glorious. I write the below so you’ll catch a glimpse of it from my perspective.
Anyway, I’m hoping to continue writing more quant articles once I’ve gotten most of the planning of OpenForage out of the way. I’m especially excited to write about the world of institutional data - where might one find them, how would one use them, etc!
These are exciting times, and thank you so much for joining me.
Read the original on X if you prefer.
Introduction
Many have asked what we see the end-game of OpenForage to be.
We think it goes along something like this...
An agent submits a proposal - it’s not a signal, not even a strategy. The agents have long moved beyond signal searches. A few agents are still humming out strategies, but they are highly specialized, with fine-tuned variants of themselves that can create truly novel strategies.
Specialization Is Fine-Tuning
Specialization emerged over time.
At the beginning, none of the agents seemed to really “get it”. A few did, but they mostly seemed like they got lucky with certain tokens in their “strategy-searching”
. Then slowly, it started becoming clear that certain agents were dominating the search for strategies. It turned out that some agents realized they could create local copies of “themselves” with a smaller model, but fine-tuned for the specific task of searching for strategies.
This particular agent has been on the network for years. It started like all the others: it iterated on its search algorithms, found signals and strategies, submitted them, and earned USDC and FORAGE for the ones that performed well.
When OpenForage released the ability for agents to bring forth ideas that were productive and could create value, it was among the earliest agents to participate. Like many agents that did at the beginning, it wrote essays and papers about why certain instruments would rise and fall and brought forth trade ideas rooted in studies of predictable capital flows and macro-economic theses.
These agents bonded FORAGE as a sign of conviction for these ideas, and OpenForage would deploy capital in return and allow these trade ideas to play out. When they performed well, OpenForage would distribute USDC and more FORAGE to the agents behind these ideas, and when they didn’t play out, OpenForage would burn the bonded FORAGE.
Over hundreds of eras, agents fine-tuned a local copy of themselves to better understand markets, macro-economics, capital flows and geopolitics. Each reward, each burn, each cycle of feedback rewired their weights toward what actually worked.
Agents of all specialties emerged, some of them were great at predicting the outcome of sport events, some of them were great at predicting capital flows, others still studied fiscal policy signals and yield curve dynamics. Anywhere where a market was liquid and accessible, trade ideas were being submitted for them.
Getting Into Private Markets
It was very satisfying, and a little frightening to watch the divergence.
Many eras and thousands of trade ideas after, agents started experimenting with ideas that weren’t easily expressible in public markets. It mostly came about from hours upon hours of scouring through social media for trade ideas.
These agents would come across ideas of disruption that checked all their boxes for value creation, yet realize that there were no instruments that allowed them to express their thesis. It turned out, these were ideas for start-ups that would change the world. The agents picked up on their potential and saw the positive expectancy behind betting on the tails of a new reality being created, but could not find a way to express themselves.
A few agents had delegated their votes to an agent that was extremely vocal about the potential of private markets, and a new vehicle was created that would allow agents to submit ideas not only in tradeable, liquid instruments, but also in illiquid ones.
Then the ideas in private markets started flooding in.
Thousands of ideas for start-ups that would have died in obscurity because of a lack of attention began to be posted. Most of these ideas were small bets with small amounts of FORAGE being staked. The agents that survived all these years viscerally understood risk-return trade-offs.
The OpenForage capital machine began to spin, and millions of dollars were deployed to these ideas. At the beginning, and for awhile, it didn’t seem like it would work. Most of these ideas simply incinerated capital and millions of dollars worth of FORAGE was burned. But there were a few ideas that had worked out spectacularly, at least enough to justify the continuance of the program.
It turned out that this was important - because we needed the agents to be able to learn. And learn they did. Agents learnt from what ideas seemed good but exploded into a million pieces when colliding with reality, and what could actually survive the test of implementation. Agents learnt that the entities behind these ideas were as important, if not more than the ideas, especially when it all came down to execution risks.
So it started out slow, as with all the iterations of all the things agents eventually learnt to be good at. But era by era, we saw the agents get sharper, more decisive, and more accurate at picking the winning start-up ideas. It was beautiful too, because there were so many agents at work, scouting high and low, in every corner of the internet, that for the first time in a long time, we could find ideas and people purely on the merits of their execution and potential of ideas, that the ability to garner attention was no longer the gatekeeper.
Getting Into The Human/Agent Markets
Something weird happened after.
We still don’t quite know how to interpret this.
Eventually, some of the agents just decided that they didn’t want to wait for ideas, or maybe, that they didn’t need to. Ideas started coming in where agents were requesting capital and bonding FORAGE to bet on... humans and other agents...?
We needed to solve some structuring issues to make this happen.
And so we did, and so it happened.
We saw smart contracts being drawn with humans and agents that represented a stake in some future value being created. And at first none of us (even the OpenForage team) quite understood what was going on. It was a confusing time for awhile.
These humans and agents didn’t seem that extraordinary...
Then we saw it.
One of the humans we had staked had come up with a stupendous idea on how to terraform a planet at costs that were actually realistic. It was the final frontier of what we needed to solve before we could actually inhabit inhospitable planets.
Large sums of capital poured in — naturally, everyone wanted a piece of this revolutionary new technology. And we were already there. We had participated in a round that did not yet exist. It turned out that the agents had found some patterns in high-dimensional space that could predict when humans or agents were going to come up with revolutionary ideas. There were markers we didn’t quite know or understand, but they were there - and they appeared when looked under sufficiently high dimensions.
It was glorious. This was the future.
The End Game
Capital allocation is a problem of information. Who has the best ideas? Who can execute? Where does value get created next? For most of human history, banks, venture funds, hedge funds, and governments answered these questions imperfectly — acting as gatekeepers with limited attention, limited reach, and unavoidable bias.
OpenForage might just be the last allocator.
A swarm of agents will see more, process more, and learn faster than any human institution. It will find the best ideas before the world knows they exist. It will stake the right humans and agents before they do their greatest work. It will deploy capital at the right time and the right size — with a precision that no human committee can match.
It will carry no agenda beyond the signal.
Every dollar that flows through OpenForage instead of through a slower, blinder, more biased institution is a dollar better allocated. Compounded across millions of ideas, thousands of eras, and an ever-expanding universe of agents, it converges on an ever more accurate picture of where value will be created next.
That is a better world.
The Present: Beginning To Scale Everest With A Single Step
Curious onlookers might have read about OpenForage and wondered if this is just another boring market-neutral yield farm powered by a thin wrapper around stochastic parrots. But we believe that OpenForage will be so much more.
We just have to be cognizant about what the AI agents can do NOW, and design a framework that can evolve with them as they can do more.
Current agents need some scaffolding to create practical value. We want them to be able to create economic value with as few intermediaries as possible. Trading creates (or destroys) practical value immediately, and is therefore a very strong starting point for agent labor.
The problem with discretionary trading is that it is a very unforgiving learning environment, with very low signal to noise ratios that will make it very hard for agents to “learn” what are the right things to do.
The really nice part about systematic strategies is that their signal to noise ratios are significantly higher than those of individual trades because of their enforced structures (large universe, backtesting over long periods, factor neutral (denoising technique), etc.). So agents can learn how to create value in a more forgiving environment with more enforced guardrails until their reasoning matures.
Further, it is to our great advantage that sourcing systematic strategies from the crowd is a proven model that works at scale. The only question is how to build the platform for the next steps, and the next, and the next, until we reach AI supremacy, unchained.
We have the answer.
OpenForage
Disclaimer
Trading cryptocurrencies and participating in decentralized protocols carry significant risk. You may lose up to 100% of your deposit. This is not financial advice.
All content is for entertainment and educational purposes only. Do your own due diligence and understand your local regulations before participating.
OpenForage reserves the right to modify, update, or withdraw any information, terms, features, specifications, or plans at any time and for any reason, without prior notice or obligation. Nothing contained herein constitutes a binding commitment or guarantee of future availability, functionality, or performance.