Introduction

Most "AI makes a game" demos still glue a chat agent onto Unity and hope the glue holds. Wanaka Studio 1.0, launched by Wanaka (@wanakaofficial) on November 2, 2026 (18:00 UTC), takes the other bet: ship an AI-native 3D multiplayer engine the agent can actually drive — models, rigging, gameplay logic, and sound in one browser workspace — then publish a playable link with nothing to install.

Primary surface: studio.wanaka.app. Marketing overview: wanaka.app. Vendor benchmark page: eval.wanaka.app. Closed SaaS, free to start. This is a same-day game-gen drop for studios watching prompt-to-playable pipelines, not another image-model headline.

The Magic: Engine Built for Agents

Wanaka's launch pitch is blunt: stop making AI fight Unity; build an engine it can use. Studio 1.0 packs five claims that matter for interactive pipelines:

  1. All-in-one workspace — 3D assets, auto-rigging, gameplay logic, and audio without hopping tools.
  2. Speed and agent cost — on Wanaka's eight-task internal eval, Wanaka Max averages about $5.74 / 23.9 minutes of agent cost versus Opus 5.5 + Unity at $32.19 / 121.1 minutes (~82% lower cost, ~80% less time). Treat those as vendor-recorded workflow numbers, not a third-party Elo.
  3. Runtime ambition — they market AAA-style visuals with 4K / 60+ FPS, real-time physics and lighting inside their engine.
  4. Direct control — edit the scene yourself; the agent continues from your changes instead of regenerating the whole project.
  5. One-click publish — every game becomes a cross-platform browser link. Publishing itself is billed as zero tokens (creation tokens are separate).

The human / agent / engine split is the design thesis: you set taste and feel, the agent implements, the engine keeps the result editable and runnable — including native multiplayer and share flows.

Pro Tip

Before you burn a Max run on a hero campaign, open eval.wanaka.app and skim the eight-task table (FPS campaign, coastal photography, arcade racing, swordplay ARPG, multiplayer challenge, plus three mini-games). Note the caveats in their own data notes: Wanaka costs are agent fees only (generation tools excluded), some Unity mini-game rows are extrapolated, and quality is their team's qualitative call. Use it as a planning budget, not gospel.

Hands-on: From Sentence to Share Link

You do not need a local engine install. Flow as marketed:

  1. Open studio.wanaka.app and describe the game in one sentence (racers, shooters, platformers, tower defense, puzzles, multiplayer arenas, and similar templates are on the table).
  2. Let the creation agent plan and assemble scenes, characters, logic, and audio; step into the visual editor when language is the wrong tool (nudge a building, retune feel).
  3. Playtest in-browser, then publish → build → share. Friends open the link on desktop or phone with no download.

Starter prompt shape that matches their demos:

Make a small multiplayer arena where two players grab glowing orbs on a floating night-market rooftop. Add a 60-second round timer, simple dash dodge, and a scoreboard. Keep the camera third-person and the art style soft neon dusk.

If you are comparing agent+engine stacks this week, park Unity/Claude glue projects beside a Wanaka Max run on the same brief and log wall-clock plus spend — that is the only comparison that survives marketing slides.

Conclusion

Wanaka Studio 1.0 is a closed, browser-first AI-native game engine aimed at playable 3D multiplayer worlds with shareable links — not open weights, and not a Comfy node. For ArtRealmAI's game-gen beat, the actionable surface is live today at studio.wanaka.app, with the launch post and cost/time claims to pressure-test on eval.wanaka.app. Watch for open export paths later; right now the story is prompt-to-playable in one SaaS runtime.

Original Source:

—Titus