That’s SWEAi. Autonomous multi-strategy crypto trading infrastructure that reads seven live data feeds, trades across five exchanges, remembers every decision it ever makes, and gets smarter every single day.
Most trading bots forget yesterday. SWEAi remembers everything. Every tick it observed, every thesis it formed, every trade it took, every outcome. Vector-indexed. Queryable in milliseconds. Before every decision, SWEAi asks itself: “when have I seen something like this before, and what happened next?”
Every closed trade updates the Memory Palace with the full context that led to it. Win or lose, SWEAi tags the setup with what worked and what didn’t. Over time, it stops making the same mistake twice — and a human never has to tell it what to stop doing.
When a new signal fires, SWEAi retrieves similar historical setups from the Memory Palace and weighs the current decision against that evidence. This is not a static model. It’s a growing, evolving record of market patterns that gets more valuable every day it runs.
The Memory Palace is auditable — every stored thought, decision, and outcome is inspectable. You can ask SWEAi “why did you take that trade?” and it will show you the memories it pulled, the regime it detected, and the thesis it formed. No black box. Ever.
A retail trader has their gut. A quant fund has their backtests. SWEAi has every one of its own past decisions, searchable at 40ms latency, growing every minute. That is the moat.
Crypto is a 24/7 global derivatives market with real, measurable inefficiencies — funding-rate arbitrage, sentiment-driven volatility compression, and cross-venue price divergences. Human traders cannot capture them at scale. Institutional quant shops can, but won’t bother below ~$100M AUM.
Crypto perps trade 24/7/365 across five+ major venues. Arbitrage opportunities appear and disappear in seconds. Human reaction time is the binding constraint — and SWEAi removes it.
Price, order book, funding rate, options IV, on-chain flow, sentiment, macro calendar — seven streams at once. A trained model can weight them all; a human picks one or two and hopes.
Most retail traders lose because they can’t resist overtrading or cutting winners early. SWEAi enforces its own position sizing and exit rules. It has no ego to protect.
That’s the gap. And we’re racing to fill it before anyone else does.
SWEAi is infrastructure, not a single strategy. The Regime Router chooses which strategy to deploy based on current market conditions; every trade is tagged with full context and fed into the Memory Palace so the system can learn which theses produce winners.
The brain. Orchestrates all modules, enforces global kill-switch and drawdown halts, writes a structured brain_log of every thought and decision, and auto-restarts any module that crashes.
Multiple strategies gated by an ADX-based regime classifier. Currently two validated strategies live (RSI Oversold 4H, Squeeze Breakout). Two more were shipped, rigorously tested, and removed when they failed walk-forward. Honesty is a feature.
Prices stream continuously via WebSocket — sub-second latency across every connected venue. The Analyst synthesizes the latest state into a market thesis on a short cycle (currently every few minutes, moving toward fully reactive). Trades get attributed back to theses so we track which kinds produce winners.
Vector database indexing every observation, decision, and outcome. The Coordinator queries it before every decision. The longer SWEAi runs, the smarter it gets — not because we retrain, but because it remembers. This is the moat.
Every strategy gated by an XGBoost classifier (64.5% CV accuracy). Quarter-Kelly position sizing. Circuit breakers at the module, account, and global level. A drawdown of 4% pauses trading; 6% triggers halt.
Five exchange adapters (DXtrade, dYdX v4, Kraken Spot, OKX, Binance). Hot-swappable — the Coordinator can decide at runtime which venue is best for a given signal. Live on dYdX mainnet today with real USDC.
Six weeks of build, 90+ commits, deployed to production infrastructure in Nuremberg, Germany. Paper trading a prop challenge on DXtrade; live with real capital on dYdX mainnet.
Three revenue lines, each targeting a different segment. The infrastructure is the same; only the capital source changes. Prop farming is the engine — fastest path to profit, self-funds the rest.
This is the core fund generator. Pay ~$150 per challenge, pass at +10% return, receive $12K funded capital paying 80/20 profit split. SWEAi operates dozens of challenges in parallel at zero marginal labor cost. Profits from passed accounts are auto-reinvested to buy more challenges — a compounding loop that reaches usable scale faster than any other revenue line. When users want to start taking profits to their pocket, they set thresholds: “once my SWEAi account hits $X, sweep Y% to my bank.” Plaid bank connection coming in Phase 2.
Deploy company capital into high-conviction setups — particularly funding-rate arbitrage, where we capture spread between venues with near-zero directional exposure. Live on dYdX mainnet today.
Productize the platform. Retail traders pay a flat fee or a cut of their profits for SWEAi to run their account. Institutional customers pay for API access to the regime classifier and signal feed. Launches after 90 days of live track record.
Seven pipelines today, five exchanges today. Here’s what’s next and what each piece costs to build.
Cost: ~1 week engineering per pipeline + $50–500/mo in API fees for paid data. Total: ~$15K to ship all five plus 12 months of API runway.
Cost: ~2 weeks engineering per adapter. Total: ~$20K to ship all four.
Public-facing app at app.suncitysys.com. Users sign up, connect their exchange accounts via read-and-trade API keys (withdraw disabled), pick a risk profile, and SWEAi trades their account. Plaid bank integration for automated profit sweeps. Multi-tenant auth, per-user isolation, full audit trail.
Cost: ~$40K engineering + $5K infra + $10K legal/compliance review. Total: ~$55K to launch private beta to 100 users.
Four additional strategies already coded but not live (Grid Scalper, EMA-MACD, Mean Reversion v2, Pairs Trading). Each needs WF+MC validation before regime-router wiring. Revive Fear Extreme & Trend Rider with new gates once we have richer live data.
Cost: ~1 week per strategy re-validation. Total: ~$15K.
Hire a part-time risk analyst to review every strategy change and produce monthly risk reports. Independent audit of backtest methodology. Legal formation work for the SaaS tier.
Cost: ~$40K / year.
$30K seed into prop challenge farming (enough to run ~100 challenges in the first wave). $100K into dYdX/Kraken proprietary funding arb, where larger positions reduce relative fee drag materially.
Cost: $130K capital at risk (not operating expense — capital).
Operator and system architect. Previously shipped quant trading infrastructure for the cryptosnyper project. Owns the roadmap, risk framework, and capital deployment decisions. Sun City Systems LLC is the corporate entity.
SWEAi itself is built in tight loop with Anthropic’s Claude models, which act as a pair-programmer and research analyst. Every code change runs through a 327-test automated suite before deploy. The velocity advantage is structural: a one-person shop ships like a five-person team.
We are raising $350K to deploy capital into prop farming and proprietary arbitrage, ship the user-facing app, and extend the platform with four more exchanges and five more data pipelines.
We’ll send the full deck, term sheet, financial projections, and read-only dashboard credentials within 24 hours.
Prefer a direct email? invest@suncitysys.com