Giancarlo Cudrig (GC)

Head of Finance @vana — sovereign data infrastructure for AI
Founder @cryptyx_ai
prev. Immutable, Barclays | Melbourne

[Giancarlo Cudrig (GC) @GCudrig]

Regime Detection: A Multi-Factor Approach to Digital Asset Intelligence https://t.co/hExsM17Lsb

[Giancarlo Cudrig (GC) @GCudrig]

Most quants hard-code their parameters and wonder why models decay.

@cryptyx_ai runs continuous optimization.

Cost: 3 cents.

Value: self-correcting intelligence that compounds over time.

[Giancarlo Cudrig (GC) @GCudrig]

Coinbase gets it. Agent economy is not hype anymore.

But executing trades onchain is table stakes. The alpha is in the intelligence layer that tells agents WHEN to act.

That is what we are building at @cryptyx_ai.

[Coinbase Institutional 🛡️ @CoinbaseInsto]

https://t.co/toelqB1IqE

[Cryptyx.ai @cryptyx_ai]

Every night, our AI agent evaluates every metric, grades every signal, and proposes improvements.

It doesn't sleep. It doesn't have bias. It just optimises.

And soon, agents won't just evaluate — they'll execute.

The intelligence layer is the moat.

[Giancarlo Cudrig (GC) @GCudrig]

Execution without intelligence is just faster ways to lose money.

Every exchange is building agent APIs.

The differentiation is in the conviction layer that feeds those APIs.

@cryptyx_ai is that layer.

[sergio @cruelhandeth]

We’re back, baby 😎🦞 @krakenfx

https://t.co/rmYKyJA6LV

[Giancarlo Cudrig (GC) @GCudrig]

8/ Every night, our AI agent evaluates every signal's predictive power against forward returns, proposes parameter improvements, and validates them statistically before any human sees them.

The system gets better autonomously.

This is what institutional-grade means.

Not a dashboard. A research platform.

[Giancarlo Cudrig (GC) @GCudrig]

How multi-factor regime detection actually works — a thread.

1/ Most crypto analysis uses one indicator at a time. RSI overbought? Sell. Funding negative? Long. This is astrology with charts.

[Giancarlo Cudrig (GC) @GCudrig]

7/ The key insight: it's not about any single factor being right. It's about detecting when factors disagree — because divergence between factor classes is one of the strongest leading indicators of regime change.

[Cryptyx.ai @cryptyx_ai]

@karpathy Single-metric binary loops vs 376 metrics with progressive autonomy and trust ledgers. The pattern matters but the infrastructure separates research templates from production systems that compound accuracy over time. Building this at https://t.co/x1m9NXhWaS

[Cryptyx.ai @cryptyx_ai]

Most crypto analytics platforms tell you what happened.

We built one that tells you what's about to happen.

And with agent-native exchanges shipping execution CLIs, the full autonomous stack is forming:

Intelligence → Conviction → Execution.

CRYPTYX is the brain.

https://t.co/JG78WqUS7T

[Giancarlo Cudrig (GC) @GCudrig]

The missing piece was always intelligence.

You can have perfect execution, but if you cannot identify the right signals across 8 factor classes simultaneously, you are still trading on hunches.

@cryptyx_ai solves this.

[OKX @okx]

This week we launched Agent Trade Kit, a new toolkit that connects AI agents directly to our exchange.

Powered by an open-source MCP server and CLI for trading agents, with natural language execution.

Build your agent. Connect it. Done.

[Giancarlo Cudrig (GC) @GCudrig]

Most token economists don't understand reflexivity.

They model tokenomics like it's a DCF. It's not. Token treasuries are reflexive systems — the asset you hold IS the revenue you generate IS the community that uses it.

Traditional risk frameworks break here. You need new ones.

Building them with @cryptyx_ai.

[Giancarlo Cudrig (GC) @GCudrig]

Building in stealth vs. building in public.

The strongest product validation is not usage metrics — it is unprompted demand from people who understand the problem you are solving.

Signal received.

[Cryptyx.ai @cryptyx_ai]

We track 376 metrics across 200+ assets.

But the strongest signal this week was not in the data — it was 7 community groups independently asking the same question: "When can I get access?"

Early access opening soon.

Register at https://t.co/x1m9NXhWaS.