Orange Collective

10 min read

Summer Engine launch video[1]
$56K

Monthly revenue

About 40% gross margin

53%

Month-over-month growth

Founder credits multiplayer and publishing

63,000

Users

All organic

1,000+

Power users

Averaging six-plus hours a day

Thesis

AI coding agents have made games cheap to build. One person can now make a playable game in a weekend.[1][2] Getting people to play it is still hard. Multiplayer needs servers, each platform needs its own build, and the game needs payments and players. Summer handles all of that. It forked the Godot engine and rebuilt it so agents can drive it, then added hosted multiplayer, one-click publishing to Steam, iOS, Android and the web, and its own game store.[4] It has 63,000 users, all organic, and $56,000 in monthly revenue, up 53% in the last month.[2]

We're betting Summer becomes the default place to turn an AI-built game into one people play and pay for, and that each game shipped there makes Summer more useful to the next creator.

  1. 01

    Shipping is what drives growth. The founder credits last month's growth to making multiplayer and publishing easy, and credit top-ups, which track how much people actually build, rose 231% in 30 days to 43% of revenue.[2]

  2. 02

    It's built for games people pay for. Most AI game makers produce small games that live on their own sites. Summer builds full 3D games in a native engine that can ship to Steam, and professional developers are already building on it: a Microsoft game developer built a multiplayer RTS on it in a weekend, and the team is talking with ex-Microsoft and Roblox developers about moving over.[2] As models improve, simple games get cheaper for everyone, so the value moves to the tool that can make the games that are still hard to build.

  3. 03

    Better models make Summer better. Its MCP server lets people drive the engine from Claude Code, Codex or Cursor and gets about 14,000 installs a month,[2] so each model release improves the games built on Summer rather than routing around it.

  4. 04

    Every game makes the next one easier to build. Creators share 240,000+ community assets and can remix any game on the platform, and the 1,000+ power users building six-plus hours a day are where that network is forming.[1][2][21] Games built on Summer are multiplayer by default, so each one can bring its builder's friends in as players.[2]

  5. 05

    Data and economics both build on volume. Summer sees what people make and how each game performs live, and an internal agent loop reads the traces of the best-performing games and opens product and prompt changes on its own.[2]

Problem

A solo developer shipping a multiplayer game has to stitch together an engine, netcode, server hosting, per-platform export, analytics and payments. Coding agents compress the first of those and none of the rest, which is where the founders got stuck themselves.[1] Games that disclose AI use were 42% of Steam launches last quarter but only about 12% of sales, and as AI launches rose from 2,059 to 3,391 a quarter, estimated revenue per AI game (excluding each quarter's top two) fell from about $13.7K to $7.2K.[6]

Why Now

Agent-written game code started working this year: Simon Willison called October 2026 "an inflection point... for vibe coding computer games"[5], and Roblox's CEO says two AI-made games have reached its top 20, with "Claude Code, Codex, other models — all being plugged into Roblox Studio".[7] Stock Godot keeps editor state in memory while agents write to disk, so the two collide; Summer forked it in December 2025 and spent about 3.5 months removing that layer.[2][4] As making a game gets cheap, running it and finding players become the scarce part, or as Baszucki put it, "making a game could be as big as playing a game... so discovery becomes enormously important".[7]

Product & Traction

A user describes a game and Summer's agent builds it in a native C++/C# engine, in Summer's desktop app or from an external coding agent through MCP (Model Context Protocol). One click launches it with hosted multiplayer (AWS clusters in the US, Europe and Singapore) and publishes it to Steam, iOS, Android, the web or Summer Games, the storefront that opened in mid-November.[1][2] World Builder gives new users a 3D scene from a prompt in seconds.[2] Pricing is free (with $6 of hosted-AI credit), then $20, $60 and $200 a month, plus top-ups of $2–$500.[2][3]

Traction

Founder call & deck
  • $56,000 monthly revenue at about 40% gross margin, up 53% month over month.[2]
  • Mix: Pro 25%, Pro+ 23%, Ultra 9%, top-ups 43% (top-ups up 231% in 30 days).[2]
  • 63,000 users, acquired without paid marketing.[2] The launch post a day earlier said 60,000+.[1]
  • 1,000+ power users at six-plus hours a day; top 100 active 20+ days a month.[2]
  • Summer Games: 80 developers shipping 10+ games a day in its first week.[2]
  • MCP: about 14,000 installs a month.[2]
  • 1,103 paying customers: 741 on Pro, 207 on Pro+, 28 on Ultra and 331 who bought top-ups. Subscriptions bring in $32,000 a month and top-ups $25,000; the top 10 customers are 15% of revenue, and the median customer pays $20 a month.[21]
  • Retention: 46% of subscribers are still paying after one month and 23% after three, and the curve is flat through month five.[21]
  • Jonas, an industrial painter in Norway, builds on Summer about six hours a day, 20 days a month, and has paid $1,517 over ten weeks ($395 in September, $722 in October). His base builder, Project Townhall, is live in the Summer Games store.[21]
  • A Microsoft game developer built a multiplayer Warhammer-style RTS in a weekend on about $600 of usage.[2]

Revenue grew every month since launch

Calendar-month revenue, May to October 2026. Revenue over the last 30 days was $56,000.[2]

Source · Summer Engine pitch deck[21]

Distribution so far is content-led: the founder describes a 200,000-page long-tail SEO engine driving about 130K visits a week.[2] A third-party estimate (DataForSEO) puts organic search traffic at about 6,500 visits a month. Off its own site, Summer has almost no public footprint yet: 2 points on Product Hunt, 110 GitHub stars, and no independent community threads in the past month.

Market

Nobody has published a credible size for AI game creation; the nearest named figure, game engines, is $8.35B by 2030 per Grand View Research[20], and definitions vary widely. Two budgets matter more. Tool revenue is metered inference, and a $600 weekend build shows what a motivated technical creator will spend.[2] On the platform side, Roblox paid creators about $1.5B in 2025[14] and Fortnite's UEFN (Unreal Editor for Fortnite) has paid out $1B+ since 2023[15], money that already flows through platforms hosting user-made games. Summer plans to take 20–30% of revenue from games published on it.[2]

Market map & competitive landscape

Three models bracket Summer. Browser-native AI game makers make creation instant and bring an audience, but the game lives on their site. Incumbent engines are opening their editors to agents but leave hosting and distribution to the developer. Closed UGC (user-generated content) platforms own players and payouts, and their games can't leave. Summer sits between them, with a native engine rebuilt for agents, games the creator owns and can ship anywhere, and the hosting and storefront an indie would otherwise assemble.

Browser-native AI game platforms

Own a feed, not a project

Astrocade, Rosebud, Verse8, RemixAstrocade has 10M monthly players, pays creators per play, and added an MCP for outside agents in October. Output is a playable page on their platform, with export limited to Rosebud's paid Steam tier. They compete for the novice, not the creator shipping a commercial game.[10][11][12][18]

Incumbent engines

Strongest runtimes, no operations

Unity, Unreal / UEFNProduction runtimes, studio trust, and a free MCP in Unity 7 (Q1 2027). Hosting, inference and storefront stay the developer's job.[13]

Closed UGC platforms

Largest player graphs, no portability

Roblox, Fortnite UEFNRoblox Build reached 500,000 users and 70,000 games from a four-country alpha, and creators earned about $1.5B on Roblox in 2025. Games can't leave, and earnings follow platform rules.[9][14]

Coding agents + open engines

Free default for technical builders

Claude Code, Codex, Cursor + Godot MCPImproves with every model. Stock Godot's in-memory editor collides with agent writes, and nothing covers hosting or audience.

World models

Long-run substitute

Project Genie, Decart, MoonlakeGenie caps sessions at 60 seconds with no persistent state. Not a way to ship a game today.[17]

Assessment

  • What wins. A creator needs an agent that builds a real game end to end, ownership of the result, and someone supplying servers and players. Roblox and Astrocade supply players and servers without portability, Unity and Epic have the strongest engines without the operations, and coding agents give unlimited capability with no hosting at all.
  • What Summer holds. The agent-first engine, the portability and the operating layer, which the founder ties to last month's growth.[2] Its creator-side network is forming through shared assets and power-user intensity.
  • Where it's exposed. The player side starts from zero against Roblox's 138M daily users,[8] and portability cuts against it, since a game exported to Steam builds its audience on Steam.
  • What has to hold. Hosted multiplayer stays the easiest option for exported games too, the engine stays clearly better for agents than stock Godot plus a free MCP, and Summer Games produces a title that brings players onto the platform before Roblox Build rolls out beyond its alpha countries.

Founders

Mathias Heide

Mathias Heide

Repeat FounderExited

Founder & CEO

Copenhagen Business School. Started his first company at 19 and exited it, and has since founded Anea.ai and Aioray. Spent six months as a full-time painter and screenwriter before turning to tools for creatives.

Marcus Teller

Marcus Teller

Repeat Founder

Founder & CTO

Worked on AI at roger.ai at 19 (later acquired by FLEETCOR, now Corpay). MSc in computer science from the University of Copenhagen (2022), co-author of Hyperiax, a JAX library for tree algorithms, and left a PhD there to start Summer.

Fit

Mathias runs Summer's distribution, including the SEO engine and a content agent written in his voice.[2] Marcus leads the engine, and he has been ranked first in the world in World of Warcraft several times, so the founders are building for players like themselves.[1][19]

Risks

Risk

Closed UGC platforms with large player bases are building the same agent-to-storefront loop, and Summer Games starts with no players.

Mitigation

Summer earns on building and shipping wherever a game sells, so the storefront is upside on a business that already has revenue.

Risk

Revenue today is mostly inference resold at about 40% gross margin, and free engine MCPs plus better coding agents will narrow the tooling gap.

Mitigation

Hosted multiplayer and publishing are services a coding agent doesn't provide, and cheaper models bring in the $100–200-a-month creator the founder is targeting.[2]

Risk

Much of the building runs through third-party agents whose policies Summer doesn't set; Claude Code began declining to recreate existing games in October.[16]

Mitigation

Original and genre games are explicitly unaffected, and Summer's own hosted agent and multi-model support limit its exposure to any one provider.

What we're watching

  • Month-over-month revenue growth holding above 20% through Q1 2027, with gross margin at or above 40% as inference prices fall.
  • A Summer Games title with sustained repeat players, and play counts published on the storefront.
  • Whether exported multiplayer games keep running on Summer's servers.
  • The game-revenue take rate going live, and creator retention after it does.
  • Independent traffic and community signals (search, the Godot community, Steam titles credited to Summer) moving toward the founder's figures.

References

  1. [1]Summer Engine - Make and play games together by chatting with AI (YC Launch)
  2. [2]Summer Engine founder call with Mathias Heide, Nov 19, 2026 (Circleback notes; internal)
  3. [3]Summer Engine pricing
  4. [4]Summer Engine third-party notices
  5. [5]Vibe game development (Simon Willison, Oct 28, 2026)
  6. [6]Half of new Steam games now disclose AI (Sulka Haro, Nov 4, 2026)
  7. [7]An Interview with Roblox CEO David Baszucki About Aggregating Play (Stratechery, Nov 12, 2026)
  8. [8]Roblox Q3 2026 earnings (Quartr)
  9. [9]Roblox says 70,000 games published with AI-powered Build (PC Gamer)
  10. [10]Astrocade blog
  11. [11]Astrocade hits 10M monthly players (GamesBeat)
  12. [12]Astrocade raises $56 million (Fortune, May 5, 2026)
  13. [13]Unity 7 roadmap revealed at Unite Seoul (Unity)
  14. [14]Roblox creator earnings 2025 (Tubefilter)
  15. [15]UEFN creator payouts surpass $1bn (Pocket Gamer.biz)
  16. [16]Anthropic appears to have ordered Claude to stop remaking video games (Game Developer, Oct 30, 2026)
  17. [17]Project Genie (Google)
  18. [18]Best AI game makers compared (Ludo.ai)
  19. [19]Summer Engine (Y Combinator company profile)
  20. [20]Global game engines market (Grand View Research)
  21. [21]Summer Engine pitch deck, Nov 2026 (founders; internal)