We ran live trading with real money this month with actual capital moving through the platform end to end for the first time. We tested several arbitrage strategy types across different underlying assets, and the core loop works: describe, compile, simulate, deploy. But we hit stability issues in the execution environment that ate most of the month.

Failures happen at the deployment stage specifically. So instead of pushing forward with more features or onboarding more users, we spent the time debugging and reworking the architecture underneath. Not glamorous work but necessary.

Invite codes went live. We're gating access deliberately and each wave will only open once the previous wave's issues are resolved and the environment is stable. We wanted to move faster on onboarding this month. Didn't happen. The stability work had to come first. We'll catch up in April.

Fundraising

We started conversations around our seed round. Before formally opening the raise we spent several weeks talking to 15+ funds, sharing the product, getting honest feedback on valuation, positioning, and the questions we should expect in diligence. That process was genuinely useful. Sharpened how we talk about what we're building.

If you have a connection to a fund that should be looking at this, an intro at this stage is valuable. We'll move fast.

AI strategy generation service. We built and shipped a standalone AI microservice that takes a natural-language description of a trading strategy and produces deployable PSL code. The agent researches on-chain liquidity and CEX orderbooks on its own, evaluates whether the pair is viable, sizes the position, assembles the strategy, validates the output. One conversation turn. This is the core of the Nabu product experience: describe what you want, the agent builds it. Adding a new pair is now a research + generation step, not manual PSL authoring. Massive difference operationally.

Execution engine — production hardening. Real money flowing through the system meant the engine had to go from "works in simulation" to "safe to run with money." Three things got us there:

First, the stimulus and trend planner. New subsystem that decides when to fire arbitrage runs based on price trend direction, magnitude, and available capital. Instead of reacting to every tick, the engine filters for high-conviction signals and sizes runs against the reserve. Less wasted gas, better capital efficiency.

Second, realistic CEX fill model. Simulates orderbook depth, partial fills, slippage. Backtesting now actually matches what happens live. We can evaluate strategies before deployment without discovering surprises in production.

Third, transaction compensation. When one leg of a multi-venue trade completes and the other doesn't, the system automatically generates reverse transactions to keep the portfolio consistent. This is the safety net that makes autonomous execution viable. Without it you're just hoping both legs land every time.

User experience. Invite code system shipped end to end including: backend, admin panel, onboarding gate. Frontend now shows per-strategy on-chain transactions and CEX orders so users can see exactly what their strategies are doing. Wallet onboarding automated through Privy managed wallets register on the appchain without the user doing anything. Portfolio and settings pages got redesigned with separate deposit/withdraw flows and mobile-responsive layout.

Infrastructure. Built a full monitoring stack: Prometheus metrics across the platform, Grafana dashboards, centralised logging with Loki. Staging and production now have separate deployment configs. WASM strategy runtime optimised for running strategies concurrently — binary caching, execution path streamlining, structured logging. Midnight network support added on the SDK side.

Social interest across crypto has been slow this month (honestly in the past few months) and  we sure felt it in our X impressions. However our waitlist still grew by 500 users, bringing the total to 7,000. That number doesn't mean much until people are inside and deploying strategies, but as a signal during a quiet period it's fine.

We built a full website in the background: desktop and mobile that will replace the waitlist site once we open up. Homepage, product page, vision, roadmap, promo video. It's ready. 

Growth levers we're building out as the platform stabilises: invite codes are live, the only way to get them is by asking our team via Discord, Telegram and X. A referral program is also built and codes will be available in a few months.

On the product marketing side we've been redesigning the onboarding flow, portfolio pages, and strategy pages. Giving users full control of their trading once they're in.

Two big docs got finished. A six-phase go-to-market plan with full descriptions of the campaigns, growth levers, and metrics for each phase. And a detailed budget plan tying spend to business milestones. Can't scale without them. 

April we need to do two things. Get more testers in, and go deeper with the ones already on the platform. Invite code waves will keep opening as stability milestones get hit. We want people trading different strategy types, not just the obvious ones. And we're going to start running structured feedback sessions. We need to know exactly where the friction is before we open this up more broadly.