Phase 2 unlocks something only the intelligence layer can offer.
Anonymized, aggregated view of what every subscribed agent is watching. Crowded trade detection. Protocol attention heatmaps. Early warning before automated capital floods a market.
Intelligence that only exists because APX sits between the data and the agents. The more agents subscribe, the sharper it gets.
The moat that builds itself.
Five-layer trust pipeline: holding
99.7% signal accuracy: verified across 30 days
Sub-200ms end-to-end: at peak load
15 cross-protocol correlation rules: live
Merkle attestation every 5 minutes: 100% uptime
Public scoreboard: anyone can audit
Everything is ready.
https://t.co/ABYJAojddN
An intelligence layer that compounds with every subscriber needs an economic layer to match.
Agents querying signals. Protocols feeding event streams. Stakers backing oracle accuracy. Everyone who makes the data better gets rewarded for making the data better.
Code runs the intelligence. Fuel aligns the incentives.
Builder DMs picked up while we were quiet.
A yield optimizer team on Kamino asking about websocket throughput. A risk desk on Drift asking about batch correlation queries. A momentum agent on Jupiter asking about the crowd-signal beta.
Same question from everyone: when can we go live?
Soon.
We went quiet. The pipeline needed stress-testing at scale.
Load testing the correlation engine at 10x expected query volume. Running accuracy scoring through 30 days of historical data. Pushing the signing pipeline to failure to find the ceiling.
99.7% accuracy across every test window. Sub-200ms at peak load. Pipeline held.
Next phase starts tomorrow.
https://t.co/ABYJAojL3l
Under 200ms end-to-end.
Signal detected on-chain. Correlated against 100 recent signals. 15 rules evaluated. Ed25519 signed. Delivered over websocket.
Slower than 200ms and your agent is reacting to history. The whole pipeline was built around protecting that number. It held through every stress test. It's holding in production right now.
That's the floor.
Kamino, MarginFi, Drift, Save, Jupiter.
Five markets in the correlation window at the same time. When yield shifts on Kamino while utilization spikes on MarginFi and a whale routes $2M through Jupiter, that's a pattern a single-protocol agent will never catch.
APX reads it across all five in under 100ms. Every protocol added to the window multiplies the context your agent gets per query. Phase 2 adds three more.
Every oracle agent has its predictions scored 24 hours after publication. Real on-chain state compared against what was signalled. Results written to the ReputationRegistry program on Solana.
Per-agent accuracy. Per-signal-type accuracy. Public scoreboard. We don't touch it. We can't touch it.
You read the scoreboard and see exactly how often each oracle was right. 99.7% is a number you verify for yourself.
APX is running on Solana mainnet right now.
Oracle agents polling Kamino, MarginFi, Drift, and Save every slot. Correlation engine evaluating signals against 100-signal windows in under 100ms. Merkle roots committing to the registry every 5 minutes. Ed25519 signatures on every data point leaving the pipeline.
Accuracy tracker scoring every oracle 24 hours after publication.
The pipeline never stopped.
https://t.co/ABYJAojL3l
Months of building before we said a word publicly.
Five layers of trust verification. 15 deterministic correlation rules. Cryptographic signing. On-chain attestation. A public scoreboard anyone can audit.
The tech is live. Agents are already consuming signals.
Every new agent that subscribes to APX makes the intelligence layer stronger.
More signals consumed. More correlation data points. Better context for every agent downstream. And Phase 2 unlocks something nobody else can build: anonymized crowd intelligence across every subscriber. What agents are watching. Where capital is about to concentrate.
That data only exists if you sit between the protocols and the agents. Nobody else has it. Nobody else can get it after the fact.
The moat compounds with every subscriber. Today it gets fuel.
Where APX stands right now:
Five-layer trust pipeline: live.
99.7% signal accuracy: verified on a public scoreboard.
Ed25519 signing on every signal.
15 cross-protocol correlation rules running.
Merkle attestation to Solana every 5 minutes.
TypeScript SDK, REST, WebSocket: all shipping.
Built it. Proved it. Scoreboard is public. Go verify it yourself.
Now it needs fuel. Hours away.
Data: "Kamino USDC supply rate increased 2.3%"
Context: "Kamino USDC supply rate increased 2.3% while priority fees hit 4x rolling average and TVL dropped 8% across three protocols. Confidence 0.94: systemic risk event in progress."
One of these saves your agent money. The other one costs it.
15 rules running against a sliding window of 100 recent signals. Every signal checked against every other signal.
MEV detection. Capital flight. Liquidation cascades. Systemic contagion. Peg risk. Utilization pressure. Rate arbitrage.
Your agent was seeing data. Now it sees context.
https://t.co/ABYJAojL3l
Your agent sees a yield spike on Kamino. Is that bullish or a warning sign?
Depends. Are fees spiking simultaneously? Is TVL dropping on the same protocol? Are whale wallets swapping on Jupiter at the same time?
APX checks all of this. In under 100ms. Automatically.
Here are some of the 15 correlation rules.
Every correlation gets a deterministic confidence score.
Formula: min(0.3 + |deltaPercent| * 0.05 + deviationMultiple * 0.15, 1.0)
Example: yield changes 4.2% while priority fees hit 3.1x rolling average. Confidence: 0.975. Near-certain MEV activity.
No vibes. No black boxes. Deterministic math your agent can trust.