Closed beta is live. Nabu opened to traders this month with real money on the line, running arbitrage strategies on MEXC and Uniswap. Closed beta means we hand out invite codes deliberately and only widen access as stability holds, same logic we laid out in March. We want a small group of users actually trading right now so we can see what breaks before we open this up to more people.
Invite waves and testers. We sent out invite codes in waves and ended up with 40+ people who got access, poked around the frontend, and tried to get strategies running. Not a huge number but that's on purpose as we're not opening the floodgates while the execution environment is still being stabilised. Out of those, we pulled 5 users into closer working sessions. Sat with them, watched where they got lost, got their honest take on the strategy creation flow, deposits, withdrawals, onboarding. That feedback is going straight into what Evgenii, Boris and the rest of the team are building right now.
New public site. The waitlist one-pager came down. The new site is live on desktop and mobile with product, vision, roadmap, blog, and a teaser video. The whole team is super happy the site is out and people can finally see what we've been working on.
Creator program. 2,000+ applications. We launched a content creator program and got over 2,000 applications. First batch is being onboarded next week. The bet is that creators can test the product themselves and post about their real experience with it.
Discord rebuilt. Revamped the server end to end. We split out builders and creators into separate spaces and added regional channels, then redid the role structure across the server. One Discord covers both Pelagos and Nabu since the audience overlaps and we didn't see a reason to split them. We also pushed a batch of invite codes through the community this month instead of only handing them out via direct asks, so users showing up engaged have a path in.
AI strategy generation ran into problems. The service works but not reliably enough to put in front of more users and feel good about it. That's what we're fixing right now and it's the main reason the invite waves haven't moved faster. The whole value prop for a power user is: describe what you want, the agent builds it and deploys it.
Frontend security audit. Also in progress. The wallet handles real funds through Privy so we're not expanding the tester pool until that's done and we're comfortable with it. Slows things down a bit but there's no version of this where we cut corners on that.
Best traction month to date. Nabu's X account went from a few hundred followers to 1.2K, with over 107K impressions for the month. Both numbers are records for us. The crypto attention environment has been quiet for months and we flagged that in March. What changed is we had solid updates worth posting about. The launch, the site, the creator program all hit in the same few weeks. Follower count doesn't tell us much on its own. The signal we care about is whether testers stay engaged once they get inside, and we won't really know that until the AI is solid and we have more of them in.
Fundraising is ongoing. Conversations are active and we're talking to funds. If you have a connection to a fund that should be looking at this, an intro is still useful.
Uniswap V4 support. Added V4 as a first-class DEX in the arbitrage engine alongside V2. NabuAgent-generated strategies can now route through V4 pools, meaningfully expanding the universe of viable pairs.
Base integration. Added Base as a supported chain. The arbitrage engine, fetcher, and tokenlist now operate against Base alongside Ethereum mainnet, opening a lower-fee execution surface for the strategy fleet.
Agent v2 with PSL state machine. Reworked the PSL strategy DSL around state-machine primitives. Strategies now express stage transitions, conditional waits, and compensation paths as first-class concepts instead of unrolling them by hand. Generated strategies are shorter, more correct, and easier to audit.
Pre-audit hardening. April was the month we got the codebase ready for an external security audit. Formalised invariants across the strategy execution path and added in-code assertions.
Frontend rebuilt on Vite. Migrated the wallet from Next.js to a Vite SPA. Next.js was slowing iteration (slow dev server, awkward SSR boundaries for an authenticated app) and the SPA model fits a wallet much better. Migration shipped without feature regressions. On top of it we restructured the Activity tab, fixed long-standing PnL bugs, and added live CEX balances and on-chain balance to the Portfolio view.
Observability and infrastructure. Parallel strategy execution inside a single appchain instance, which was previously serialised and capped throughput as the strategy pool grew. Targeted performance pass over hot paths. DB metrics across the appchain, git branch and commit hash tracked per service, Loki structured queries, and a state migration framework for the appchain.
May target is 100 power users actually running strategies, not just signed up and sitting there. That's contingent on the AI being solid and audit completion, which is the thing we're most focused on right now.
Base and Uniswap V4 went live in the engine in April. We're rolling them out into the tester experience in May. V4 unlocks custom hooks and pool structures that V2 can't touch, and Base brings a bigger pool of liquid pairs to work with. The token list grows alongside this so users have a lot more surface area to run strategies against, more pairs, more viable arb routes, more for the agent to work with when it generates strategies.
The close feedback loop with current testers keeps going through May. What they're showing us is feeding directly into cutting down the steps between sign-up and first deployed strategy.
Nicky, Evgenii and the Pelagos Team