We quietly built something powerful at ForgeX and made over $1+ Million in profit.

While the rest of the market chased headlines, we turned our attention to DAO forums. Our question was simple:

Could governance proposals reliably predict market movement?

So, we trained an AI agent — not just to skim proposals but to understand sentiment, forecast impact, and trigger live trades based on high-confidence insights.

We didn’t launch a product. We didn’t raise capital. We just built internally. We scraped over 6,500 proposals, trained models to process them, and created a trade execution engine to act on high-signal opportunities.

That system went on to generate over $1+ Million in profit. Today, we’re making the core open-source for the community to explore, use, and build on.

Governance Moves Markets

DAO proposals aren’t background noise. They’re decision-making signals. Protocol changes, treasury allocations, fee structure tweaks, and validator incentives — all these things show up in governance forums before they ever hit token prices.

But most people ignore them. We saw them as opportunity.

These proposals were early indicators of price action. So we trained a model to interpret sentiment, score the impact, and trade accordingly. That’s where the real edge was hiding.

Our System, End to End

We built a complete flow — from raw text to executed trades — without any social media inputs or human guesswork.

First, we scan governance forums like Aave, Compound, Sushi, and Lido in real time. The Proposal Scanner picks up new posts, cleans the text, and logs them to our database.

Next, we process the proposal through our fine-tuned RoBERTa sentiment model. Trained on more than 6,500 DAO-specific proposals, it assigns a sentiment score. Only proposals with a score of 0.80 or higher, and validated as genuine, are eligible for action.

The trading logic then kicks in. If the signal is bullish, a long position is triggered. If bearish, a short. But only if market conditions are stable — BTC price volatility is checked before any trade is approved.

If approved, the execution module places a futures trade on Binance using $200K capital and 3× leverage. It sets a 2% stop-loss and calculates a take-profit range based on the model’s confidence.

What the Trades Looked Like

Below is a detailed look at real trades triggered by the governance agent. Each was initiated based on a high-confidence sentiment score (≥0.85), filtered through our decision logic, and executed with clearly defined risk parameters.

Over the test period, the agent took 13 trades in total. Not all were winners — some were closed at breakeven, and a few at calculated losses due to volatility filters or post-entry reversal signals. The goal wasn’t to maximize every move. It was to systematically capture asymmetric upside, while managing downside with controlled risk and predefined exits.

Total Profit: ~ $1.5 M

Disclaimer: This is not financial advice. The system was tested internally with strict risk parameters. Past performance does not guarantee future results. Always do your own research.

A Few Standout Trades:

Uniswap — Protocol Fee Adjustment
Uniswap’s proposal to increase gUNI’s allocation to 2% of swap fees didn’t generate much noise publicly. But under the hood, it pointed to higher LP rewards and an increase in yield-generating activity — both positive signals. Our agent flagged it at a sentiment score ≥0.85, entered the trade at $9.58, and exited at $12.28 within 4.7 days. That move returned $169,027 on a $200K position. Structural fee changes like this don’t just boost protocol economics — they often front-run token valuation when caught early.

Aave — Launching Emergence (LRT Consolidation)
Aave’s proposal to merge Etherfi and Lido’s LRT flows into a single initiative, “Emergence”, was one of the more complex governance proposals. But the implications were powerful: consolidation of TVL and increased protocol-native utility. The agent entered at $197.85 and exited at $258.25, locking in $183,193. This trade captured both the liquidity narrative and the protocol moat in one move.

QuickSwap — Launch on Sony’s L2 (Soneium Minato)
When QuickSwap was announced as the first DEX partner on Sony’s Soneium Minato, the headline sounded more like PR than a trade signal. But our model picked up on the credibility and reach Sony brought to the deal. The agent entered the position at $0.04499 and exited at $0.05839 in under two days — earning $178,813. Institutional alignment often takes time to reflect on-chain, but in this case, price moved immediately.

Synapse — $600K/Month Grant with Roadmap
At first glance, a grant proposal to fund Synapse Labs with $600K/month didn’t sound like a reason to long. But our agent paid closer attention to the roadmap bundled into the grant — which included bridge upgrades and RFQ routing enhancements. The trade was entered at $1.1893 and exited at $1.536 just 24 hours later. Result: $175,201 in profit. Development funding that implies short-term utility improvements is an underrated trading catalyst.

QuickSwap — Expanding Falkor Perps to Ethereum
The proposal to extend Falkor perpetual contracts to Ethereum didn’t draw headlines — but it did open up a large addressable market for QuickSwap. Perps expansion is capital-intensive and demand-driven, making it a strong signal for engagement growth. The trade opened at $0.03197 and closed at $0.0404 in just over a day and a half, yielding $158,401 in profit. Not all scaling proposals show up in TVL charts immediately, but they often show up in price.

Trading With Guardrails

We didn’t just let the system run wild. Every trade had to meet multiple conditions before being executed:

• Sentiment score must be at or above 0.80
• No more than 4 trades live at a time
• Never trade the same token in parallel
• Stop-loss automatically placed at 2% from entry
• Take-profit calculated dynamically
• Every action alerted via Telegram/Slack

The result was a system that stayed controlled, auditable, and stable — even while actively trading.

Why We’re Sharing This

We didn’t set out to build a product. We built an engine — one that others can now remix.

The Governance Agent is open-source because we believe the infrastructure layer for agent-based systems in Web3 shouldn’t be closed. It’s ready to be used as-is or extended into something new.

You can train your own models, point it at different blockchains, change the trade logic, or build an interface on top.

Explore the code: Governance Trading Agent

What’s Next?

While this system is out in the open, we’re not done. Behind the scenes, we’re experimenting heavily — across trading, DeFi, DAO analysis, NFT markets, and more.

Some of these ideas might turn into new agents. Some might never see the light of day. But we’re running a wide set of experiments to find the next edges in on-chain data.

We might release the tooling we’ve been using to build these systems. We call it the ForgeX SDK — a modular framework that powers these agents under the hood.

If and when the time is right, we’ll open it up. In the meantime, if you’re curious and want to build with us…

Join the Waitlist

We’re opening up early access to the SDK, tools, and playground.

You can join the waitlist here:
https://forgexai.com

Team ForgeX