May was our first full month with users running strategies on live funds, and most of it was spent sat close to that first group of testers. We’d set a target going in of 100 power users actually running strategies by the end of month, and we didn’t get there. The reason isn’t a surprise, it’s a choice. We don’t widen access until we’re ready for it on our side, and on two specific things we weren’t quite ready yet.

What May actually became was the work behind that. We pushed the strategy agent forward, sped up the execution engine, got the external audit done, and used what testers showed us to cut friction out of the path they take through the product. On the go-to-market side this was our strongest month yet. The details below.

The AI generation problem from April is fixed. This was the main thing slowing our invite waves last month. The AI strategy generation now does buy/sell and simple arbitrage reliably, which is the bar we needed it to clear before putting it in front of more people. We gave it to about 10 users this month and sat close with them, watching how they used it and where it still felt rough. The whole value prop for a power user is to describe what you want, the agent builds it and deploys it. That now works for the strategy types a new user reaches for first.

Two things still gate wider access. First, the AI fees are higher than we want them to be. The service works, but each generation costs more than we’re comfortable putting on users right now, and we’re bringing that down. Second, the audit. The external review was carried out this month and we’re working through the findings now. We’re not opening access wider until both are closed. That’s the honest tradeoff behind missing the May number. We’d rather be a few weeks late than open up before either is right.

X ads are cleared, and the first paid test worked. We resolved all our issues with X ads, so the account is now ready to launch paid campaigns. We ran our first test and got impressive results, over 2.5M impressions and thousands of link clicks to the website, all on a limited test budget. Organic impressions came in around 120K, our best month ever. Follower count stayed flat at 1.2K. Like we said in April, follower count on its own doesn’t tell us much. What’s promising is that paid is now a lever we can actually pull, and the early numbers say it’s a good one.

A new UI for the Portfolio and Strategy pages. Following our first power traders’ feedback, we designed a complete new UI for the Portfolio page and the Strategy page. It goes live in the next few weeks.

Pricing is designed. We also designed the full pricing strategy for Nabu and the UI that goes with it. It ships once the product is ready for market, but the thinking and the design are done, so it won’t be a scramble when the time comes.

The newsletter moved to Beehiiv. We moved our newsletter from Mailchimp to Beehiiv. We’d built up to 7K subscribers on Mailchimp when they decided to stop our service. Their policy toward crypto isn’t the friendliest. So we had to move, and we went with Beehiiv, which we’re really happy with. They’re crypto-friendly, the UX is great, and our open rate on the first newsletter jumped to 19.94%, which is high. We can’t wait to use it more often to share updates with our subs.

The creator program. Applications closed this month. We got over 2,700 applications and the team is reviewing every single one. We've already shortlisted a few good candidates, and we'll be onboarding them slowly as we see good fit.

KOL pilot. We also worked with a small group of KOLs this month. They built trading strategies with Nabu themselves and posted honest run-downs of how it went, the good and the rough. Around 25K impressions between them and hundreds of reactions. The bet is the same one behind the creator program: creators testing the product and posting about their real experience lands better than anything we say about ourselves.

Website traffic. Traffic on nabu.pro was relatively quiet early in the month and then jumped from mid-May onward as campaigns started to kick in. About 1,168 page views and 831 active users month-to-date, with an average session around 59 seconds.

With real money running through the platform, May went into the parts of the product that have to be solid before we widen access. Mostly that’s the strategy agent and the execution engine, plus the guardrails around both.

A stronger strategy agent. We made a round of improvements to the agent that turns a plain-English request into a working strategy. Before it writes anything now, it researches the market. It looks up the token, checks the pools and liquidity, sees what’s actually tradable. So the strategy it builds actually matches what’s there. It also checks its own work against a growing list of known mistakes before handing a strategy back, catching errors that used to slip through.

Faster, more reliable execution. On the engine side, we finished the work to speed up the path from a strategy spotting an opportunity to an order actually filling, and built the monitoring to measure every step of it. The picture from production is encouraging. Most of the price data a strategy reads is fresh to within a tenth of a second. And it earns its keep: it let us catch a live strategy that had silently stalled instead of trading, the kind of issue that’s invisible without it. We also made orders behave correctly under real conditions. Partial fills, retries, stale prices, all handled cleanly now.

Tested with real users, on real money. This was the first full month with testers running strategies with live funds, and we worked closely with that first group. Watching where they got stuck, then feeding it straight back into the agent, the engine, the onboarding flow. We also stood up automated end-to-end tests across the wallet so the path our testers rely on doesn’t quietly break as we keep shipping.

Base is ready. Base is fully built and integrated into the engine. It’s a lower-fee chain that gives strategies more pairs and cheaper execution. It’s done on our side. We’ll switch it on for testers once the audit findings are closed.

Audit done. We finished audit preparation and the external security audit was carried out this month. Working through the findings is the main item for next month, with the fixes shipping inside the same window. That’s the gate that unlocks wider access.

Hyperliquid integration. We’ve already started building the Hyperliquid integration and it should go live in June. That brings perps into Nabu for the first time. New surface for the agent to work with: basis trades, funding rate plays, perp/spot arb across venues.

Uniswap V4. Bringing V4 into the tester experience. V4 unlocks custom hooks and pool structures that V2 can’t touch, and the token list grows alongside it. More pairs, more viable arb routes, more for the agent to work with when it generates strategies.

Improving the agent and the user path. We keep pushing on the agent. Sharper market analysis, more reliable strategy generation. And on the path the user takes around it, cutting down the steps between sign-up and a first deployed strategy. The close feedback loop with current testers keeps going through June and feeds directly into both.

Post audit fixes. Working through the findings from the external audit and shipping the fixes. Closing this out unlocks broader access.

June is about finishing the audit and the AI work and then opening up. Alongside that we keep working on AI efficiency to push the fees down. When the audit is closed and the fees are where we want them, we open access to a much bigger group and start pushing for real volume. The 10 testers we have now keep going through June, and what they show us keeps shaping the onboarding and strategy creation flow.

— Nicky, Evgenii and the Pelagos Team