The problem GoalGuard solves

What is GoalGuard?

GoalGuard helps everyday ETH holders protect money they have set aside for important real-life goals, such as rent, tuition, travel, or emergencies, without requiring them to understand options trading.

The Problem

Most crypto holders think about what their money is for, not about strikes, premiums, expiries, or complex trading strategies.

When ETH prices fall, money reserved for an important expense can suddenly lose value. Existing options platforms can help manage this risk, but they are often too technical for everyday users.

How GoalGuard Helps

Users describe their goal in plain language:

“I have $1,200 in ETH for rent next month, and I cannot afford to lose more than 5%.”

GoalGuard then:

1. Converts the goal into clear financial limits.
2. Searches live Thetanuts options for suitable downside protection.
3. Shows the cost, coverage, expiry, and possible outcomes in simple language.
4. Uses three independent Gonka reviewers to evaluate strategy, risk, and consumer clarity.
5. Generates the exact unsigned Base transaction preview for inspection.

Why It Is Safer

GoalGuard is designed to keep the user in control:

• Financial calculations are deterministic rather than invented by AI.
• Three independent AI reviewers assess every plan.
• Any rejection, uncertainty, missing data, or stale market information stops the process.
• Users see the risks and trade-offs before any transaction.
• No private keys are sent to the server.
• The hackathon version never requests a signature, broadcasts a transaction, or moves funds.

Why It Matters

GoalGuard makes on-chain options more understandable and approachable. Instead of asking users to become derivatives experts, it connects protection directly to the purpose behind their money.

In one sentence: GoalGuard turns “I need this ETH for rent next month” into a transparent, independently reviewed downside-protection plan, making on-chain options easier to understand and safer to explore.

Challenges we ran into

Challenges We Ran Into

1. AI Responses Were Not Always Valid JSON

One specific bug occurred when some Gonka-hosted models added a <think>...</think> reasoning block before the requested JSON response. Even though the model’s answer was correct, our application could not parse it, which caused the entire review to fail.

We solved this by:

• Removing reasoning blocks before parsing.
• Extracting only the JSON object from the response.
• Validating every response against a strict schema.
• Giving the model one controlled retry when its output is malformed.
• Blocking the protection plan if the response still cannot be verified.

This made the three-reviewer council much more reliable without blindly trusting unpredictable AI output.

2. Live Options Data Can Change Quickly

Thetanuts orders can change or expire between the moment a plan is reviewed and the moment its transaction preview is generated. Using an outdated quote could show users incorrect costs or unavailable protection.

To overcome this, GoalGuard re-fetches the selected order before creating the preview. It compares the latest order with the reviewed version, checks liquidity and deadlines again, and creates a fingerprint of the quote. If anything has changed, the plan is marked as stale and the user must review fresh information.

3. There Was No Safe Testnet Execution Path

The Thetanuts integration for this track uses Base mainnet, so testing a complete trade could involve real funds. We did not want the demo to expose users or judges to unnecessary financial risk.

We therefore designed the hackathon version to use live market data and the real transaction-building process while stopping at an unsigned preview. The application shows exactly what would be sent on-chain, but it never requests a signature, broadcasts a transaction, or claims that the user is protected.