Hi everyone,
August was the month we put the mobile client in front of the most new players it has ever had. Our second Limited Market Test brought close to 8,000 new players from Brazil and Latin America. We shipped guest user on iOS, replaced the old multiplayer server with our own headless Bevy build, launched the Friendzone Buildathon, and the Treasury Mandate with Avantgarde started executing.
Scope: Aug 1–31
Living Roadmap: high-level goals, % complete, and short descriptions across hiring, projects, and funding.
Financial Dashboard: treasury and ops overview, diversification activity, vault deposits/yields, and spend categories.
Financial Transparency
MANA price on 8/31: $0.0713.
Q2 Transparency Report. The Q2 2026 report, with the attestation from our community auditor, is published here: DCL Regenesis Labs Transparency Report Q2 2026.
DCL Regenesis Labs Wallet 
DeBank Dashboard
| Date | Assets | Amount | Price | SubTotal |
|---|---|---|---|---|
| 8/31/26 | MANA | 10,738,209.00 MANA | $0.07130 | $765,634.30 |
| 8/31/26 | Smokehouse USDT | $419,651.63 | $0.9990 | $419,231.98 |
| 8/31/26 | Smokehouse USDC | $359,458.46 | $1.0009 | $359,781.97 |
| 8/31/26 | Gauntlet USDT Core | $295,437.57 | $0.9990 | $295,142.13 |
| 8/31/26 | Gauntlet USDC Prime | $276,519.96 | $1.0009 | $276,768.83 |
| 8/31/26 | Morpho | $5,382.45 | $5,382.45 | |
| Current Total Assets | $2,121,941.67 |
Operational Wallet DeBank Dashboard
| Date | Assets | Amount | Price | SubTotal |
|---|---|---|---|---|
| 8/31/26 | MANA | 1,361,398.34 MANA | $0.07130 | $97,067.70 |
| 8/31/26 | USDC | $139,839.52 | $1.0009 | $139,965.38 |
| 8/31/26 | USDT | $7,500.00 | $0.9990 | $7,492.50 |
| 8/31/26 | ETH | 0.0194 ETH | $2,458.99 | $47.70 |
| Current Total Assets | $244,573.28 |
Treasury Mandate Wallet DeBank Dashboard
On June 24 the DAO moved 339 ETH and 20,000,000 MANA from the DAO treasury into this Safe, to run the covered-calls strategy with Avantgarde that we described in the June update. August was the first month with activity: the ETH went into a Morpho WETH vault, and the first covered call was written on 3,741,394.79 MANA at a $0.0668 strike, expiring September 5. The net premium, 12,720.74 USDC, went into a Morpho USDC vault.
| Date | Assets | Amount | Price | SubTotal |
|---|---|---|---|---|
| 8/31/26 | MANA | 16,258,605.21 MANA | $0.07130 | $1,159,238.55 |
| 8/31/26 | MANA (covered call collateral) | 3,741,394.79 MANA | $0.07130 | $266,761.45 |
| 8/31/26 | WETH in Morpho WETH vault | 339.36 WETH | $2,458.99 | $834,480.50 |
| 8/31/26 | USDC in Morpho USDC vault | $12,752.55 | $1.0009 | $12,764.03 |
| 8/31/26 | ETH | 0.051 ETH | $2,458.99 | $125.41 |
| Current Total Assets | $2,273,369.94 |
Grants Program: Season 1
In August, several Season 1 projects were landing scenes in the market test (more on that below). That gave us real numbers on how well each one keeps new players.
Candy Sheep Meadow had the best day-one retention of any landing scene in the test. Its creator kept running events, which brought in a community of players from Asia.
Flag Tag had the longest median playtime of the test (5.3 minutes) and is fully multiplayer.
Adventure Land (Space Runner) and 
PixelArcade were both in the rotation and closed further milestones.
Doge Hunt went through a stress test with our team, ahead of its final review.
Season 1 closes in September, and Season 2 is confirmed for later this year.
Mobile Client
:android: Android · 
iOS v1.12.0 to v1.12.2
40 PRs merged. We shipped 1.12.0, 1.12.1 and 1.12.2 during the month.
Guest user on iOS (shipped)
Guest user went out on Android in July, and in August it shipped on iOS with 1.12.0. Players on both platforms can now start without signing in. We also added an in-game notice and badge that invites guests to upgrade their account (#2622).
New HUD and social panel
- First iteration of the HUD and Social Panel redesign (#2642), followed by a second iteration at the end of the month (#2723)
- A dedicated HUD layout for Preview Mode, so creators testing a scene see what they need (#2686), and a scene stats panel (#2663)
- HUD overlays no longer swallow touches meant for the scene’s own UI (#2616)
Loading and stability
- Fixed two cases where the loading screen could stay up for minutes or forever (#2640, #2641)
- Sign-in deep links now complete on a cold start (#2689), and content requests send a User-Agent, which stopped a set of CDN errors (#2660)
- The market test brought in a much wider range of phones than we test on, many of them below our minimum spec. 1.12.2 added better crash reporting so we could find the root causes, and stability became the main focus going into September.
Avatars, emotes and multiplayer
- Pulse, the new transport for avatar positions, is in the client (#2541). More on Pulse in the Bevy section.
- Upper-body emotes, so a player can wave while walking (#2690)
- Avatar LOD based on a pool, with billboards only when a scene is crowded (#2700), and smoother movement for other players’ avatars (#2778)
Creator tools & SDK
- Light sources for scenes, with limits that keep them safe on mobile (#616)
- Billboard
targetEntitysupport (#2507) and Cast v2 video streams (#2625) - Preview hot-reload, so creators see changes on their phone without restarting the scene (#2720)
- We wrote the documentation for custom mobile controls (
TouchScreenControlsandUiInputBinding), rewrote the safe-area page aroundscreenInset(#172) and refreshed the mobile Missing Features page (#178)
From ad to scene
At the end of the month we merged the change that sends a player who installs from an ad straight into the scene that ad showed (#2788). This is what lets us advertise a specific game instead of Decentraland in general.
In-App Purchases
The mobile Shop now sells credit packs, with new art and the signing fixes behind them (#2608). Credits are already live on iOS and can be used in the Foundation’s store.
Limited Market Test 2.0
LMT 2.0 ran from July 17 to August 20, Android only, with Google Ads in Brazil and Latin America (about two thirds of the players came from Brazil). Every new player landed in one scene, and we rotated that landing scene nine times, with about 700 players per scene, to compare how each one performs.
What we learned:
- Getting players in works. Cost per player came in under target, and game entry went from 25% to 72.5%, mostly thanks to guest user and a simpler onboarding.
- Keeping them is the gap. Day-one retention doubled, but it is still half of our goal and far from the 25% that top mobile games reach.
- The landing scene matters a lot. Day-one retention ranged from 10.3% (Candy Sheep Meadow) to 4.8% (our own Chill Zone). Flag Tag and Space Runner followed Candy Sheep. The scenes that did best were light, loaded fast and had an obvious goal.
- Guests did better than registered players on every KPI (7.5% vs 3.7% day-one retention). 93.6% of the players stayed as guests.
What comes next: stability on lower-end phones, a short onboarding inside the first scene, push notifications on day 1, 3 and 7, and language localization.
Bevy Web Client
99 PRs merged, 18 of them on the multiplayer server.
The multiplayer server is live
In July we said we expected to finish our own multiplayer server in August, and it is done. Scenes that need a referee (scores, rounds, shared state) now run on a headless build of the Bevy client instead of Hammurabi, the separate Babylon.js engine we had to maintain until now. It went live around August 25.
- Published to npm and released through CI (#1077, #1082), with render-only work stripped out and a smoke test in CI (#1056)
- Tweens run on the server (#1084), each scene room keeps its own player presence (#1102), and a broken scene.json can no longer crash it (#1018)
Compared with Hammurabi on the same scene: memory went from 308 MB to 175 MB, startup from 1.5–3 seconds to 0.7 seconds, dependencies from 62 packages to 2, and disk from 236 MB to 154 MB. CPU use for a single scene went up (6.7% to 16.5%).
Pulse: avatar positions off LiveKit
In busy scenes like Flag Tag, players were falling out of sync at 15 or 16 people, because LiveKit was dropping packets once everyone’s position went to everyone else. We moved avatar movement to Pulse (#919), and by the end of the month Pulse was the only channel for avatar state (#1137). We expect this to move the limit from around 15 players to around 100. We also contributed to the Pulse server itself (pulse #39).
Changelog
Security. We treat every scene as potentially hostile, and August had a long list of limits on what a scene can do:
- Storage APIs and BroadcastChannel removed from the scene sandbox (#1051, #1043), and scene code can no longer reach the raw runtime ops (#1079)
- Caps on WebSockets, comms rate, CRDT data and RPC calls per scene (#1031, #1033, #1085, #1087)
- A watchdog for scene workers that stop responding (#1007) and a bound on LiveKit audio decoding (#1035)
Performance.
- Idle scenes stop paying a per-frame cost (#1001), scene UI got faster (#1000) and identical avatars share materials (#1006)
- The web build compiles the engine while it downloads and keeps the compiled code cached (#1009)
Features.
The Bevy preview inside Creator Hub is available as an opt-in, while Babylon stays the default.
Governance
Wearables and emotes submission fee
The proposal passed, and on August 10 we kicked off the implementation with the Foundation teams that own the Builder, the curation flow and the contracts. The fee goes from $100 to $25, with the Foundation covering $20 for the first six months, so creators pay $5. Items go through automated validation before the fee is charged, so a creator can fix issues and resubmit without paying again. The curation committee goes from 14 members to 3, focused on IP and similarity flags. We are building the open-source validator. The work is in progress, with mid-October as the target.
Points of Interest
Two POI proposals passed this year (January 13 and March 25), and neither could be applied, because the POI registry can no longer be updated reliably. Since the system needs to be replaced anyway, a draft proposal is being prepared to retire the fixed POI list and move to live discovery: clients would recommend places based on current activity, visitors, likes and events, with Genesis City scenes and Worlds in the same system. It will go to a vote once the draft is final.
Games & Content
Our content work in August was mostly about the market test: giving it good landing scenes and measuring them.
Marsh Colony (35 PRs)
The pet game we called My Dear Pet in July is now Marsh Colony. In August it got egg hatching and growth, a breeding system, pet moods, a care center, a food mini-game and a first UI revamp (#139, #124, #129). We planned the MVP for the end of August. It took longer, and the game is still in development.
Cozy Farm (16 PRs)
Cozy Farm was the landing scene with the most players in the test (1,009). We added English, Spanish and Portuguese with a language picker (#206, #208), and fixed quests and saving based on what we saw in the multiplayer test sessions (#211, #213).
Dead Surge (8 PRs)
New models for the zombie shooter (#356) and fixes after the latest SDK update.
Chill Zone (6 PRs)
Our social landing scene for the test, with a sitting system, an emote dispenser, and seat locks (#25, #24, #29). It had the lowest retention of the rotation, which showed us that new players need a clear goal to stay.
Tower of Madness (4 PRs)
UI scaling on mobile with the SDK’s isMobile and virtual resolution (#176).
Comms & Community
- Friendzone Buildathon. We launched on August 14 with a live kickoff and AMA, on DoraHacks. The prize pool is $8,000 in MANA, and workshops ran through the build phase (mobile UX and controls, performance and more), all recorded and public. We also held a WIP meetup with builders and an X Space on August 27, “When Digital Communities Become Home”.
- Community Meetup #12 on August 5.
- Weekly Friday playtests continued across our titles and the Season 1 scenes.