Will Wright's next project is a mobile social simulation called MyMems that uses generative AI to turn spoken memories into a placeable world. In a November 18, 2026 interview with The Verge, the SimCity and The Sims creator said the goal is to "take your autobiography and turn it into a toy world." Gallium Studios targets a 2027 launch.
For builders, MyMems is less a frontier model drop than a consumer pipeline experiment: speech in, structured world objects out, with privacy toggles and shareable multiplayer spaces. explainx.ai tracks Gemini Tasks on phones and Roblox AI creation tools in the same lane—on-device or cloud generation wrapped in a social product loop.
What MyMems does, according to Gallium and The Verge
Cozy game world builder metaphor for memory pieces placed on a shareable map
The Verge describes MyMems as a social simulation where:
- Players record short spoken memories on a phone.
- The system generates pieces—buildings, props, characters—you place on a map.
- AI-powered avatars can stand in for people from your life.
- Over time Wright wants the planet to read like a "map of your psyche," referencing Inside Out as a metaphor.
- Worlds can be private or shared.
Gallium executive producer Virginia McArthur called it "a cozy world builder with a really deep social simulation underneath." Product manager Anthony Maitz stressed the tone is play, not a formal legacy archive—"not meant to be your legacy diary."
Why Wright thinks Sims players were already doing this
Wright told The Verge that The Sims players often tried to recreate real families inside a toy simulation. MyMems is the deliberate version: start from your stories instead of generic suburban lots.
He also described opening a family scrapbook where people and events jump off the page until you have a "circus" of life moments—chaos by design, not a polished narrative RPG.
How Gallium answers the anti-AI-game backlash
Game controller beside scale weights for AI authorship debates in consumer sims
The gaming industry spent 2026 fighting over generative AI in awards, GTA VI statements, and indie developer surveys. The Verge notes Take-Two said generative AI has zero part in GTA VI at Rockstar, while other publishers push phone-first AI creation.
Gallium's counter-message in the interview:
| Concern | Team response in The Verge piece |
|---|---|
| Slop | McArthur: "We are not making slop." |
| Kids hate AI | Wright says his three sons are pretty negative; he calls art/authorship worries well-founded |
| Cheating with AI | Maitz says bit-by-bit player authoring stops it feeling like you cheated—his daughter who hates AI got hooked on a character creator test app |
| Scribblenauts pattern | Team compares on-demand object generation (e.g., generating a high school) to Scribblenauts-style creativity, not wholesale asset replacement |
explainx.ai covered generative AI in surveys at work and Z-Library fake citations as parallel trust fights—MyMems will live or die on whether players believe their words stay theirs.
Competition: Inzoi, Tomodachi Life, and The Sims shadow
The Verge places MyMems beside Krafton's Inzoi and Nintendo's Tomodachi Life: Living the Dream, but notes none match The Sims' pull as an EA franchise.
| Title | AI angle (high level) |
|---|---|
| The Sims | Decades of player-authored family soap operas; EA still core revenue |
| Inzoi | Unreal-based life sim with user-generated fashion and worlds |
| Tomodachi Life | Quirky Nintendo social sim, limited AI marketing |
| MyMems | Memory-native inputs and generative placement as the core loop |
Builders shipping UGC platforms should watch whether MyMems publishes APIs or stays a closed mobile loop—Wright did not announce mod tools in the Verge piece.
Privacy, data, and what is not confirmed
The Verge interview is rich on creative intent but thin on infrastructure:
- No on-record detail on which models or hosts process voice clips on November 18, 2026.
- No retention policy quoted beyond the private vs shared world toggle.
- No price or platform exclusivity beyond mobile positioning.
Until Gallium publishes privacy policy and terms, treat voice memories like any other cloud AI feature: assume upload, inference logging, and moderation are possible; read docs at ship.
For biometric-adjacent concerns—voice prints, likeness of friends—see our surveillance and camera law archive as background, not as legal advice.
Platform and business questions still open
The Verge piece does not answer several go-to-market questions product teams will ask in 2026:
| Question | Status on November 18, 2026 |
|---|---|
| iOS vs Android first | Not stated; described as mobile |
| Subscription vs premium buy | Not stated |
| User-generated asset marketplace | Not stated |
| Moderation for shared worlds | Not stated beyond optional sharing |
| Training on user voice | Not stated—assume upload until privacy docs ship |
Studios copying the pattern should draft AI nutrition labels early: which modalities (voice, image, text) leave the device, how long clips persist, and whether friends' likenesses require consent flows. EU AI Act transparency and state biometric laws are not game-industry quirks anymore—they apply the moment you scan faces or voice prints.
Comparable AI-native games and tools in 2026
| Product | What it tests | MyMems difference |
|---|---|---|
| Roblox AI creation | Mass UGC with guardrailed codegen | Memory-native inputs, smaller scope |
| Project Genie-style worlds | One-shot environment gen | Persistent psych map metaphor |
| Inzoi | Life sim customization | Less autobiographical framing |
| Google Veo art fill-ins | Institutional archives | Consumer self-story |
explainx.ai covered Veo synthetic memories in museums—institutions worry about authenticity; MyMems leans into toy authenticity instead.
Content safety and family accounts
Wright targets casual play, but spoken memories can include third parties who never opted in. Gallium's share toggle is not a DMCA policy.
Operators should plan for:
- Report flows on shared worlds (harassment, impersonation).
- Age gates if AI avatars depict minors.
- Local-only modes for schools or therapists—if Gallium ever courts edtech, FERPA enters chat.
None of that appears in the Verge interview; flag as open questions for 2027 beta readers.
What studios can prototype without Will Wright's budget
Indie teams can de-risk the same design loop with off-the-shelf stack pieces:
- Speech-to-text on device where possible.
- Structured JSON memory objects (people, places, emotions tags).
- Text-to-3D or sprite APIs with human placement in a 2D map.
- Private link sharing before public feeds.
That is essentially AI studio panel publishing meets Blender MCP—smaller scope, faster ethical iteration.
Press timeline and what to watch before 2027
| Milestone | Why it matters |
|---|---|
| Gameplay trailer with HUD | Proves loop is fun, not only novel |
| Privacy policy + DPA | Unlocks EU and school pilots |
| Closed beta signup | First real latency and moderation data |
| Platform rating (ESRB/PEGI) | Determines AI disclosure labels |
Until then, cite The Verge for creative intent, not for technical architecture.
What AI builders can learn from the design
Even if you never ship a game, MyMems encodes three systems lessons:
- Structured output beats raw chat. Memories become placeable entities, not endless dialogue—similar to how Gemini Robotics needs action tokens, not paragraphs.
- Player-in-the-loop defeats slop claims. Each authoring step is a human checkpoint, much like human-in-the-loop phone agents.
- Cozy framing lowers stakes. A toy autobiography sells better than "AI psychotherapy"—important for app-store policy and parental reviews.
Workshop and education angle
Schools and AI literacy nonprofits sometimes ask whether memory games belong in curricula. MyMems is commercial, but its design thesis—structured reflection instead of raw chat—mirrors classroom exercises that predate LLMs. If Gallium publishes teacher mode, expect debates like South Korea exam cheating reforms: who owns a spoken answer, and can proctors detect whispered prompts?
Until 2027, treat MyMems as R&D signaling from the Sims lineage: player-authored worlds still beat fully automated open worlds for trust, even when generators fill in props.
Bottom line
MyMems is not a model benchmark—it is a product bet that stepwise human authoring can make generative fills feel legitimate. Watch Gallium privacy docs, not only trailer hype, before you cite it in AI-in-games syllabi.
Indie developer checklist if you clone the loop
Before you ship a memory-to-map prototype:
- Write consent copy for third-party names spoken aloud.
- Default worlds to private; require explicit share.
- Log model version per generated prop for DMCA disputes.
- Keep human delete on every generated object.
- Run red-team sessions with teens—Wright's interview assumes family-safe tone but not technical moderation.
Ship a paper FSM (finite state machine) for the authoring loop before you add frontier models; most failures will be UX, not FLOPs.
Search and press hygiene
Journalists and SEO teams should prefer the slug will-wright-mymems over older Proxi codenames unless Gallium redirects officially. Mixed naming confuses app-store search and trademark screens. When you embed The Verge quotes, attribute Jay Peters and Gallium Studios speakers separately—Wright's sons' skepticism is family anecdote, not a product safety claim.
Related reading
- Google Veo synthetic memories in art
- Gallup Gen Z and AI companionship
- Americans negative AI sentiment polls
- Blender MCP for 3D workflows
- What is MCP?
Creative quotes and feature descriptions follow The Verge, November 18, 2026. Gallium Studios product pages did not load in our automated fetch on the same day.
