CEO @ NEIGHBOR | Building DreamCore — AI-powered platform where anyone can create, publish & play games instantly from their phone | Based in Tokyo. /ja @notf
Joined March 2023
- Tweets3.4K
- Following1.6K
- Followers1.7K
- Likes3.6K
Pinned Tweet
We let anyone use AI to instantly create 3D or 2D games entirely on their smartphone by simply describing them in chat, publish them to the world, and let players play right away in their browser.
Steam’s pre-release game reviews are taking up to 12 business days. Valve’s target is 3–5.
Valve says submissions have increased significantly, and much of the delay comes from reviewing games again after they fail the first check. It hasn't attributed the increase to AI.
On October 22, I wrote about Steam being flooded with near-identical games. My point was simple: what has been happening on YouTube and TikTok is arriving in games.
Game platforms face two choices.
Accept that future and strengthen their recommendation systems. Or take the Netflix approach and work only with selected companies.
Steam sits awkwardly between the two.
Now its review backlog is growing.
My prediction: the app-store model built around approving each submission before publication will collapse.
The choice ahead is not simply how to review everything faster. It's whether to embrace an abundance of creators—or deliberately limit who gets to publish.
Source: [Valve’s review update](steamcommunity.com/groups/st…)
My earlier post: [October 22](teapaw.org/notf/status/2113237595…)
#Steam #GameDev #GameIndustry #AppStores #Platforms
Breaking from Tokyo: Epic just announced that Blueprints won't exist in Unreal Engine 6. Not phased out. Gone from day one of Early Access.
Tim Sweeney said it on stage today at Unreal Fest Tokyo 2026, here in my home city. He even apologized, because so many developers love Blueprints.
I'll be honest: Sweeney is my favorite CEO. Last month he announced a huge pivot at Epic toward AI acceleration, and the backlash was massive. Many game companies apologize the moment they mention AI, or insist it's only for prototypes. He kept pushing the industry forward under worldwide criticism.
Today showed the same thing. He apologized to the people who love Blueprints, and he still made the cut. Being willing to throw away what Unreal spent years building takes real nerve.
The original plan was to deprecate Blueprints gradually. What changed? AI.
Inside Epic, AI has raised productivity 1.5x to 4x in programming, debugging, and level design. Their target is at least 4x across development by the end of 2026, with up to 10x in many areas expected by mid-2027.
So UE6 moves gameplay to Verse, Epic's own language, and to a new Scene Graph built from entities and components. C++ stays for low-level systems like rendering and physics. Any AI model can plug in through MCP and Python. And migration itself is AI-assisted: Epic is already converting thousands of Blueprints in Fortnite.
This connects to what I wrote about Roblox: daily users of its MCP server grew nearly 5x since Q2, and I argued the real advantage with AI is being able to rebuild your foundation from scratch.
Epic just did that, to its own engine.
Blueprints were made so humans could build games without writing code. Visual nodes are friendly for people. They're awkward for an AI. A text language with clean APIs is something an AI can read, write, and convert.
Two of the biggest platforms in games are now designing for a new kind of user: a human working with an AI.
My take: the next engine war won't be won on graphics. It'll be won on how well an AI can work inside the engine.
And for creators, the skill that mattered most was knowing the tool. Now it's knowing what you want to build, and being willing to tear down the base when a better one appears.
If Epic can drop a beloved feature because AI made rebuilding cheap, what are we still keeping only because it was expensive to replace?
Source: GameMakers, "Unreal Fest Tokyo 2026 keynote report" (Nov 3, 2026) gamemakers.jp/article/2026_1…
#UnrealEngine #GameDev #GenerativeAI #AIAgents #Tokyo
Today an AI agent checked me in for a flight, picked my seat, and got my QR code. I didn't touch the airline's site.
What I didn't do: give it my password. I still had it run in my own local browser, where I was already signed in. Handing over credentials felt like one step too far.
That small hesitation is, I think, the real story of AI agents right now.
This week a post went viral in Japan. Someone asked an AI agent about hotels for the New Year holidays, and the agent called a hotel on its own. The person ended up phoning the hotel to apologize.
Now picture it from the hotel's side. A voice AI calls to book a room. Is a real person actually coming? If they don't show, who's responsible? I suspect a lot of businesses will simply let those calls go unanswered.
My read: the bottleneck for agents is no longer capability. It's trust, on both ends. The user has to trust the agent with access, and the business on the other end has to trust whoever the agent is acting for.
I build AI products for a living, with DreamCore for AI game creation and KUMODeck for publishing and running games. Both put AI between a person and a result. What I'm learning is that "it works" isn't enough. People need to see what the AI did, and stay in control of the parts that feel personal.
The companies that win the agent era won't just build smarter agents. They'll build the trust layer around them.
#AIAgents #GenerativeAI #FutureOfWork #ProductDesign #Startups
A year ago, fewer than 1,000 games a month were made on DreamCore. Now it's over 100 a day.
DreamCore is our AI game creation platform. You describe the game you want, and it gets built.
The growth isn't coming from where I expected.
More and more people are asking ChatGPT how to make a game. And ChatGPT has started pointing them to DreamCore. When that traffic jumped, our international users jumped with it. Today, roughly half of the games on DreamCore are made by people who don't use Japanese: English, Spanish, Russian, Chinese, Korean.
Nobody told these people to come. They had an idea for a game, asked an AI how to make it, and ended up with us.
That tells me something bigger than our own numbers.
The desire to make games is everywhere. It was always there. What was missing was a way in. AI is now that way in, both for building the game and for finding the tool to build it.
I believe web games are about to have a big wave. Not a niche trend, a real shift in who makes games and how they reach players.
We're early, and we're still learning. But we're going to build for that wave.
#DreamCore #WebGames #GenerativeAI #GameDev #Startups
Impressions tell you who saw your ad.
They don't tell you who cared.
With DreamCore, creators research your product and turn it into games their friends actually play.
Launch a campaign online in minutes. From $640, 5 games guaranteed.
dreamcore.gg/business/sponso…
Built this game with @bot and Opus 5.5.
Coin Collector 3D — Arena rules🪙
・You get 20 coins once, on your first visit
・Hit someone and their coins drop — whoever grabs them keeps them
・Your coins carry over and count on the ranking
#threejs
notef.dreamcore.gg/coin-coll…
DreamCore dev log, Jan 1 – Oct 9, 2026
• 9,535 commits
• 986 PRs (768 merged)
• 234+ production deploys
September alone: 3,434 commits, 302 PRs merged — about 10 merges a day on average.
Still evolving!
You know that moment when you start making a game during breakfast... and suddenly you've forgotten you're eating?
That's the future I hope for with DreamCore.
I want game creation to become part of everyday life.
new creation screen is live 👀
chat / assets / tweaks / text / history
all in one place now
go break it 😆
new in dreamcore: when someone plays your game, you can now go play theirs.
one tap from your guestbook.
feels like how the internet should work.
dreamcore.gg/
Testing automated gameplay with Jev and a local SAM. SAM turns visual information into JSON, and Jev decides what to do next.
@MetaforDevs @typesafeai
Post a prompt. Playable game.
@elonmusk
