Andy Tsen

Google’s Project Genie demo is impressive. Turning prompts into playable worlds feels like magic. But for brands, the real lesson isn’t about world-building. It’s about distribution and retention. Making content keeps getting easier. Keeping attention does not. Games aren’t just spaces you generate. They’re ecosystems built on progression, identity, social dynamics, and community. Those layers don’t appear from prompts. They come from platforms that already have culture and gravity. Procedural creation has existed for years. UGC tools have existed for years. The bottleneck was never production. The bottleneck is audience. Roblox, Fortnite, and similar ecosystems succeed because they solve discovery, identity, and social connection at scale. They don’t just host experiences. They host communities. AI can lower the barrier to creating worlds. It doesn’t automatically give those worlds meaning, distribution, or trust. For brands, that distinction matters. The opportunity isn’t “we can generate infinite content now.” The opportunity is understanding where communities already live and how to build inside systems people return to every day. Creation is becoming cheap. Culture and retention are still expensive. That’s where the real value sits. Where do you think brands will win in an AI-generated content world: ownership, or participation?

3

72% of devs believe Steam has a monopoly on PC games. Conducted by Atomik Research, 306 industry executives were surveyed across the UK and USA. 75% of respondents were senior managers or C-suite level, with 77% from studios with more than 50 employees. The study found that for most studios, Steam accounts for over 75% of revenue. In US law, you’d need to either control the production pipeline top to bottom, or hold a dominant share through acquisitions of competitors. Steam does neither. Valve hasn’t bought out rivals and it doesn’t own all the studios that make the games sold through its store. What it does own is competence, and that’s where this whole argument starts to fall apart. Almost every other storefront has had years to compete, and almost all have stumbled. This isn’t a case of Valve killing competition. The competition just keeps shooting itself in the foot. Steam spent years refining features that now define the PC ecosystem: cloud saves, remote play, Workshop mods, forums, reviews, refund policies, and Proton support for Linux users. It built hardware, an OS, and a handheld that all tie back into the same ecosystem. Most importantly, it kept the service stable, predictable, and user-focused. That level of reliability became its moat. What developers call a monopoly is often just a reflection of consumer trust. Steam is where people go because they know their games, progress, and friends will still be there in a decade. It’s not market manipulation it’s consistency. If studios think they “rely too heavily” on Steam, that says more about their alternatives than about Valve itself. No one forced them to abandon their own launchers or build weak ones. Consumers simply migrate to what works best for them. It’s worth remembering that monopolies aren’t defined by popularity or market share alone. A monopoly becomes illegal when it restrains competition. Steam hasn’t done that. It hasn’t blocked other stores, locked content, or acquired rivals. It just runs the best version of what everyone else keeps trying to make. The day someone builds a better, more reliable, more feature-rich service, Steam will finally have a real challenger. Until then, calling it a monopoly is just a way to avoid admitting the obvious: they aren’t trying to kill their competition. Their competition is almost non-existent.

22

4 Comments

Buried lede - I've been telling you about this: "Social + messaging is 60%+ of mobile time, and social IAP hit $15B (+16%). Game companies: If you are not designing for shorter sessions (or truly sticky long ones), you are designing for churn."

3

1 Comment

44% of Roblox's top 1,000 creators are already building with AI. Last month Roblox shipped agentic tools in Studio. Planning Mode reads your code and data model, asks clarifying questions, and writes an editable action plan. Mesh Generation produces fully textured 3D assets from a prompt. Procedural Models let you adjust attributes like the number of shelves on a bookcase or chairs around a table. The Playtesting Agent simulates input, reads logs, and verifies behavior on its own. Then it loops. Surfaces a problem, suggests a fix, feeds the result back into the plan. This is not a chat assistant bolted onto an editor. It is a planner, an asset generator, and a tester running in the same loop a junior dev would. Multi-agent parallel workflows are next. Coding, testing, character creation running side by side. The positioning angle is what stands out. Roblox spent a decade as the UGC platform nobody took seriously. Now it ships creator tooling Unity and Unreal do not have in the box. The 44% says creators already noticed. The bar for a single developer just moved. Again. #GameDev #AIinGames #Roblox #IndieDev #MobileGames

2

Roblox is investing in AI-driven age estimation to enable more age-appropriate interactions, specifically for users 13 and older. At Karta, we get questions about safeguarding from our clients constantly. So it is great to see Roblox advancing safety with their new age-based communication features, announced yesterday. Read all about it here: https://lnkd.in/dsdxQwcv

21

1 Comment

Origin Lab just raised $8m on the bet that games are a valuable data source for AI. Lightspeed led the round, alongside Google and various angels including Kevin Lin (Twitch) and Kyle Vogt (Cruise). Faraz Fatemi, Partner at Lightspeed Venture Partners, sums up the pitch well: "Frontier AI is moving from understanding language to understanding worlds. "That shift requires a different class of data: licensed, structured, multimodal, and grounded in interactive environments. Origin Lab is building the missing platform between the game industry and the AI labs training the next generation of world models." Anne-Margot Rodde, Co-CEO and CCO of Origin Lab, adds some colour: "AI has outgrown the data it started with. At the same time, game studios spent decades building some of the richest interactive environments in the world, but had no professional way to bring that data to market. "The problem was never supply. It was that the technical and commercial platform to access it did not exist. We built it. The only version of this market that works long-term is one where both sides of the table have a reason to be there: consent, attribution, revenue share, and usage tracking built into the pipeline from day one." They've already secured exclusive partnerships with more than 20 game publishers representing over 50 titles. We know mobile games are gigantic experimentation platforms. It will be interesting to see how this angle develops.

19

2 Comments

𝐀𝐑𝐄 𝐏𝐑𝐎𝐌𝐏𝐓 𝐄𝐍𝐆𝐈𝐍𝐄𝐄𝐑𝐒 𝐓𝐇𝐄 𝐍𝐄𝐖 𝐑𝐎𝐁𝐋𝐎𝐗 𝐒𝐂𝐑𝐈𝐏𝐓𝐄𝐑𝐒? (pt. 4) Growing users of SuperbulletAI is crazy. We’re 2 days close to hitting $20K in revenue — a small milestone, but one that proves something big: Roblox devs actually WANT AI to build games for them. (even if they try to hide it and boycotts them) Today’s video might be our craziest demo yet. 𝐅𝐫𝐨𝐦 𝐚 𝐛𝐥𝐚𝐧𝐤 𝐛𝐚𝐬𝐞𝐩𝐥𝐚𝐭𝐞 → 𝐚 𝐜𝐡𝐚𝐫𝐠𝐞-𝐮𝐩 𝐅𝐈𝐑𝐄𝐁𝐀𝐋𝐋 𝐬𝐲𝐬𝐭𝐞𝐦 …with a single prompt. All I typed was: “let’s add a fireball skill.” SuperbulletAI instantly searched the Superbullet Marketplace, grabbed a Fireball template, installed the entire Skill System framework behind it, uploaded animations + SFX to Roblox automatically, configured dependencies, fixed missing systems… and after few minutes: 🔥 A fully functional charge-up Fireball skill was LIVE. With debris effects, damage handlers, casting animations — everything. All from an empty project. 𝐓𝐡𝐢𝐬 𝐢𝐬𝐧’𝐭 “𝐀𝐈 𝐡𝐞𝐥𝐩𝐢𝐧𝐠” 𝐝𝐞𝐯𝐬. 𝐓𝐡𝐢𝐬 𝐢𝐬 𝐀𝐈 𝐛𝐞𝐜𝐨𝐦𝐢𝐧𝐠 𝐭𝐡𝐞 𝐝𝐞𝐯. If AI can generate skills, systems, FX, animations, damage frameworks, and full game loops… …what happens when developers stop “scripting” and start directing? That’s the future we’re building. 𝐓𝐡𝐞 𝐄𝐧𝐝𝐠𝐚𝐦𝐞 Sooner or later, the Superbullet Marketplace will hit 10,000+ reusable libraries — battle-tested systems, effects, skills, utilities, full game loops, NPC brains, design templates… At that scale, SuperbulletAI becomes the standard way to create full Roblox games. One prompt → one entire game. We’re getting closer every week. 🎥 Watch the full breakdown here: https://lnkd.in/g3BZdq8N

18

2 Comments

While investors and studios burn down the world of innovation by refusing any semblance of “risk” because of their fraught prior choices and overinvestment in highly confident people who didn’t deliver … Those of us who are actually still building are doing incredible things. Craig Allen is stepping directly in the fire and is 100% correct. With Infinite Realms, we are making AI accessible story world models so your AI can operate within canon of any IP. We turn franchise IP into software. That should be something… but even our innovation there isn’t interesting unless we hit KPIs. Venture has stopped being venture and instead, as Craig said, is risk management. 🙃🙃🙃

12

7 Comments

Tencent's gaming dominance isn't about building hit games, but about their smart acquisition strategy. Their "10% wedge strategy" works like a Trojan horse. An initial small investment buys them information rights and relationships, creating an unofficial "first right of refusal" that's incredibly powerful. For example, they took a small 5% stake in Ubisoft in 2018, then significantly increased their position by 2022, all while staying under the regulatory radar that comes with outright control. This approach completely changes the power balance of future funding rounds. Once Tencent can see inside your operations, financials, and roadmap, they have information that no outside investor can match. They've changed the typical investment model into something more strategic, using that initial money as just the first move in a longer plan to gain controlling influence. I was listening to an hour-long Gamecraft Podcast recently, and this came up, and it sent me down a research rabbit hole. What's fascinating is how this creates a snowball effect over time. Each successful use of this strategy gives Tencent deeper market intelligence across multiple studios. The insights from one investment inform their strategy with the next target, creating an expanding knowledge advantage that competitors can't overcome. Industry reports show that Tencent often structures these initial investments to include observer rights or preferred terms that seem harmless at first. But these standard-looking provisions give them special access to internal operations and future investment rounds, allowing them to block competitive bids when the time comes. Here's what this means for indie studios raising money: When a strategic investor wants just a small piece of your company, think about their plans several moves ahead. The terms of that first investment could determine who ultimately controls your creative vision. For other investors looking to acquire studios, the lesson is equally important. By the time Tencent shows interest in a full acquisition, it has already built significant advantages that make competing nearly impossible.

117

6 Comments

As hundreds of recently laid-off Reality Lab staff look for new jobs, it's wild to read a recent Road to VR article (link in comments) in which Meta CTO Andrew Bosworth blames "lack of focus" for their failed metaverse efforts -- when you can see a way more pressing problem, pictured in the article itself: Cartoonish humanoid avatars made to resemble the user IRL (Zuckerberg in this case). This one design choice was crucial to Meta's failure! The very most popular metaverse platforms, Minecraft and Roblox, are intentionally low-fi, immersive through their physics and responsiveness. Their blocky, whimsical avatars are similarly abstract and customizable. The same can be said for the furry and anime-themed avatars of metaverse platforms VRChat, and the hand puppet-like avatars of Rec Room -- both of which are so popular on Meta’s own Quest 2 headset, it caused the company at one point to raise the price of its HMD. This preference for non-realistic avatars relates to the core user base for metaverse platforms: People in their teens and pre-teens, who are generally still uncomfortable and unsure about their own real life identity and appearance. That problem is even more acute for teen girls and young women, still negotiating the social expectations and judgements around their real life appearance. Presenting them with a lifelike avatar to customize -- let alone an avatar that resembles who they are in real life -- is effectively asking them to shoulder even more social expectations and judgements, just in the virtual world. Zuckerberg and Bosworth, with no prior experience in game development or virtual worlds, apparently believed the Facebook model of real identities online should apply to the Metaverse.  For newcomers like them, this is a common assumption, enforced by mainstream press coverage of Metaverse events involving real real-life celebrities — for instance, Travis Scott in Fortnite, or Lil Nas X in Roblox. In rare instances like that, it does make sense for the avatar to resemble their real life owner. But this is actually the exception. Overwhelmingly, metaverse platform users do not prefer avatars based on their real real-life appearance — even when the internal tools to customize them that way exist. Virtual worlds enable the power to experiment with identity and creativity, relatively *free* from the assumptions and judgements of the offline world. (Including, sad to say, those judgements on social media.) It's why virtual worlds tend to have disproportionately large LGBT populations, and other people choosing different races, ethnicities, and genders different from who they are IRL. (Or choosing a non-human avatar.) I doubt Meta has learned this. But hopefully some of those who've been laid from Meta can apply this in the next generation of virtual worlds.

96

29 Comments

AppLovin has owned D28 optimization for two years. 𝗧𝗵𝗮𝘁 𝗲𝘅𝗰𝗹𝘂𝘀𝗶𝘃𝗶𝘁𝘆 𝗶𝘀 𝗰𝗿𝗮𝗰𝗸𝗶𝗻𝗴. D7 was table stakes. While every other network bid on a 7-day snapshot, AXON was pricing in week four. 𝗜𝘁 𝘄𝗮𝘀 𝗹𝗶𝗸𝗲 𝘁𝗵𝗲𝘆 𝘄𝗲𝗿𝗲 𝗼𝗺𝗻𝗶𝗽𝗼𝘁𝗲𝗻𝘁. Unity is attempting to close that gap. Vector D28 IAP ROAS campaigns are live. They claim 3x spent in the case study. No mention of yield yet. Just one title. But even half that result changes the auction. The real test is whether it holds in the wild. 𝗕𝗲𝗰𝗮𝘂𝘀𝗲 𝘁𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝗿 𝘀𝘁𝗼𝗿𝘆 𝗶𝘀 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻. Most networks optimize for D7. Your business lives on D365+. The further a network sees, the less they have to guess. 𝗔𝗻𝗱 𝗔𝘅𝗼𝗻 𝗰𝗼𝘂𝗹𝗱 𝘀𝗲𝗲 𝟰𝘅 𝗳𝘂𝗿𝘁𝗵𝗲𝗿. But now Vector can see further too. It still keeps the D7 users who actually mature. Now it lets other networks buy churning users. It bids against AppLovin for valuable "late bloomers". For two years, AppLovin has every advantage. 𝗡𝗼𝘄 𝘁𝗵𝗲𝘆 𝗵𝗮𝘃𝗲 𝗼𝗻𝗲 𝗳𝗲𝘄𝗲𝗿 𝘁𝗼 𝘁𝗮𝗸𝗲 𝗳𝗼𝗿 𝗴𝗿𝗮𝗻𝘁𝗲𝗱.

92

11 Comments

“44% of the top 1,000 creators on the platform are already using Roblox Assistant or third party AI tools through MCP to plan, build, and test their games. That's a pretty staggering number for a feature set that until now, has mostly lived on the margins of the development of workflow.”

4

The recordings from the Khronos 3D on the Web event at GDC are now up, including my talk on real-time GPU texture compression for the Web with spark.js. The talk covers why texture compression matters for the Web, how spark.js brings real-time, high-quality texture encoding directly in the browser, and includes two live demos. https://lnkd.in/gKfbMqhA Note, recording at the event failed, so I re-recorded it at home. #WebGPU #WebGL #glTF

35

2 Comments

Thanks, Leah Hoyer for reinforcing what is "table stakes" and what tips the scales in terms of landing and building a successful long term engagement with co-dev partners. Table stakes: Case studies that prove quality, velocity, and delivering positive ROI. Building long term relationships: These are not transactional. Yes, money changes hands when deals are closed. But the ongoing discussion involves sharing vision, finding alignment, and being open and honest about why specific projects may or may not be the right fit. I think a lot of my friends and colleagues in the co-dev / x-dev / outsourcing / whatever you want to call it world approach partner relationships this way. How do I know? Because we also talk amongst ourselves and share our experiences openly. We refer our friends we trust to partners who may need their services. Even if that service might compete with ours, but the fit isn't right or (the best problem to have) we don't have a team ready to take on an engagement. The long and short of it is: Life's too short for us to focus only on the deal of the minute. It's a long haul - made easier and much more enjoyable - by building relationships based on trust with people who are equally passionate about making great games with great partners.

5

1 Comment

Roblox shipped two changes this week and they're more connected than they look 🤔 Cross-Server Chat and Chat Summaries are now live globally. Cross-Server Chat lets players chat with 400 people across different servers of the same experience - instead of being stuck with the few dozen in your instance. Chat Summaries fixes the other problem age-gating created - the "🔒…" wall whenever players of different age groups tried to talk. Separately, the under-16 publishing requirements is easing. The original rule demanded a Plus/Premium subscription for 2 consecutive months to publish to under-16s. After a LOT of community pushback, Roblox added a third path - a one-time, refundable 1,000 Robux fee per game. Making it far more accessible for smaller and international creators. You may be thinking 'why do I care' - well, players who chat spend more. That's true across every social platform - engagement drives retention drives monetisation, so this isn't just a UX tweaks. Every brand integration, every experience, every creator on the platform just got a lift on the floor of what's possible. What we are seeing more of from Roblox is - ship strict in service of safety announcement/change, the community reacts, Roblox softens with a thoughtful adjustment, the community celebrates the softer version. The platform lands balanced between safety and growth.

36

6 Comments

Scaling art production sounds simple on paper. In reality, growth introduces friction long before it introduces speed: - Standards drift. - Decisions slow down. - What once felt clear becomes fragmented across teams and tools. The studios that scale successfully don’t rely on momentum alone. They invest early in shared understanding. Visual language. Review principles. Clear definitions of quality. This work isn’t exciting. It doesn’t show up in screenshots. But it determines whether growth strengthens or weakens a project. Consistency is what allows speed to exist without chaos. And consistency is built long before production ramps up. When scaling feels calm, you know the foundation is doing its job. At Corbax Games, we have the operational capacity to scale dedicated art teams from 2 to 20 artists within a one-month timeframe, ensuring continuity and quality throughout production.

84

2 Comments

5 companies did $3.4B. Whole industry (269,762 games) did $14.1B That's mobile gaming in Q1 2026 👇 $14.1B in IAP revenue. One quarter. Here's who captured it: ➡️ Top Publishers (IAP Revenue) - Tencent: $1.6B - Century Games: $650M - Dream Games: $425M - Playrix: $388M - FUNFLY: $382M Top 5 publishers = $3.44B, that's 24% of the entire market. Tencent alone captures more than 1 in 10 dollars spent on mobile gaming globally. ➡️ Top Games (IAP Revenue) - Honor of Kings: $471M - Last War: Survival: $381M - PUBG Mobile: $353M - Whiteout Survival: $326M - Royal Match: $312M Top 5 games = $1.84B, that's 13% of the entire market. ➡️ Top Games (Downloads) - Block Blast!: 77M - Free Fire: 66M - Roblox: 58M - Pizza Ready!: 43M - Subway Surfers: 42M 💡 The download chart and the revenue chart almost overlap. Breaking into downloads is hard, but breaking into revenue is a different game entirely. Data by AppMagic 🪄

114

7 Comments

The landscape of game development is shifting faster than ever. We're seeing a massive rise in a new kind of critical hire: the UGC Ecosystem Architect. With platforms like Roblox and UEFN leading the charge, studios are no longer just looking for level designers. They need experts who can blend technical systems with community growth. These architects build the frameworks that allow players to become creators, turning a single game into an infinite platform. At Game Recruiter, we’ve been in the trenches of game dev since the beginning. We understand that finding someone who can balance technical design with the nuances of player-driven economies isn't just a challenge: it’s a surgical operation. If your studio is pivoting toward the UGC space and you need the talent that actually understands how to scale these ecosystems, let’s talk. We don't just find resumes; we find the people who build the future of play. #GameDev #UGC #Roblox #UEFN

3