What the Network Fund is, and how it’s made.
The Network Fund captures the collective knowledge of the Internet and turns it into a portfolio — the insights of thousands of analysts, without needing to be chronically online.
Where it began
Back in 2020 and 2021, everyone was locked at home. Many people received a $2,000 payment from the U.S. government and had little to spend it on, so they began hypergambling on platforms like Robinhood.
Some people lost their money. But a surprisingly large number of people used that time to conduct deep research and discover genuinely valuable insights and opportunities in the market. It was a remarkable demonstration of the power of collective action and the wisdom of the crowd.
What changed
Until recently, participating in that collective knowledge required being chronically online. You had to constantly monitor Twitter, Reddit, Discord, and private chats; read thousands of posts; and distinguish thoughtful analysis from noise. If you were not already part of those communities — or could not spend all day following them — it was nearly impossible to act on the information.
Large language models have changed that.
The systemThe system
The Network Fund exists to capture the collective knowledge of the Internet and turn it into something people can use. Its goal is to make the insights of thousands of analysts accessible without requiring anyone to spend every waking hour online.
The system consists of a distributed network of analysts and creators. Some may be anonymous individuals sharing their research online. Others may be professional traders, hedge-fund employees, or people posting ideas they cannot act on as part of their jobs. Each contributes thoughts, opinions, and analysis for others to see.
The creators evaluated hundreds of thousands of analysts, calculating their implied portfolios over time based on their public statements, and identified those with the strongest track records over the past year. The top 1,000 analysts were selected to start the system off.
Every post, read as it lands- 1,000
- analysts on the list to start the system off
- 26
- holdings, none above a quarter of the basket
- 42
- days before a position expires unless it is said again
Oracle node operators then stream every post from these analysts each day. Large language models analyze the posts in real time, determining whether each expresses:
A buy
Refreshes support for a holding, or opens one
Support refreshed
A sell
Reduces the position, and can remove it
Support removed
No position
Most posts. A view with no security in it
Left to fade
The oracle network processes millions of posts, interprets the implied positions of the analysts, and estimates each analyst’s portfolio using market cap weighting for sizing. It then combines those individual views into one aggregate portfolio, weighted by analyst track record.
The node operators thus collectively synthesize this information into the Network Fund: a diversified basket representing the combined insights and opinions of the selected analysts. The basket is determined each day, and on the following day the fund is rebalanced to reflect the new position.
The machine, in six steps
Six layers, one basket. Scroll to open them one at a time.
- 01
The oracle layer
Reserve Oracle is proprietary Reserve software. Node operators select analysts from an onchain list, and each day every operator submits its own basket to an onchain contract. The contract takes the median of those baskets, and that is what it proposes to governance.
- 02
Extract sentiment
Configured language models read the selected analysts’ public posts and record sentiment. A buy refreshes support for an asset; a sell removes it. Without fresh chatter, support fades over time — assets do not sit stale in the basket.
- 03
Weigh the analyst
How far a call moves the basket depends on the analyst’s track record, set by how their implied portfolio performed against the whole market rather than by how much they post. A stronger record moves a position further, and recent results count for more than older ones.
- 04
Build the report
The calculated adjustments are bundled into a report, pricing included. The current target is 25 assets, and any single asset is capped at 25% of the basket for diversification.
- 05
Form the target
Each node’s report is taken into account and each asset is graded by support against the round’s opening stake. More than two-thirds of that stake means clear support; less than one-third means a clear lack of it. Anything between remains unresolved, and the round does not force a basket.
- 06
Final rebalance
The finalized target basket creates the rebalance proposal. The execution layer values current holdings, compares current and target weights, and builds a set of bounded auction limits. The proposal then enters governance, where the existing veto and execution process remains in control.
The current BSC deployment settings include a 1% minimum rebalance, a 10% maximum rebalance, a 50% auction price range, a $1 million maximum per auction, and a five-hour rebalance duration. These are governed deployment settings, not fixed limits of the general design.
Buy once. The network keeps up with what leads.
The holder’s part is one decision. Every day after that, the operators, the contract and governance run the six steps above.
How to buy- Not trading yet. 145 countries at launch, excluding the US.
- Illustrative holdings and weights. The composition, and every change to it, publishes when the fund begins trading.
- Terms and privacy publish at launch.
- Not advice. Figures can be out of date or wrong.