Once operators, always operators. We're industry veterans who break down games, apps, and digital businesses for the people actually building them. Read by 15,000+ industry insiders weekly.
Share
๐ฎ Santa Isn't Real in Gaming
Published 13 days agoย โขย 8 min read
โ
Happy Friday, friend!
Four countries in seven days, and somewhere between a boarding gate and a hotel lobby, I missed the TWIG recording once again. So I did what our listeners do and watched it. Jen, Kress, Laura Taranto, and Josh Chandley brought the insights with the fund in the latest episode. Four stories stuck with me:
Roblox beat their number, and yet the stock went into another free-fall
Block Blast, one of my favorite mobile games, has been lying to us about ads
Anthropic added two sentences that will stop copycats (?)
Xbox continued to lose people who ship
Oh, and in the midst of my travels, I made a pit stop in Cyprus, sat down with my friend and co-host Phillip Black to talk about our lives one year after migrating from Northern Europe down South. I fully understand that this is not the typical Deconstructor of Fun style "insights delivered" podcast. Please feel free to skip it. Unless you're considering relocating and want to hear what happens when you uproot your whole life and start a new one on the Mediterranean.
Today's newsletter is brought to you by Sensor Tower and Appcharge. More about them later...
Block Blast gets exposed in a viral TikTok series. Claude says that it won't help you closne game no more and Robloxโs downfall in the market continues despite solid numbersโฆ Also in the episode: more Xbox departures, the Elder Scrolls VI and Ubisoftโs interest in Arkane Lyon.
Kress's math: Roblox's stock with and without the steal games.
Roblox reported Q3 bookings of $1.7 billion, down 12% and above the $1.58 to $1.65 billion it guided in July. The stock fell 18% the next day. Beating a bar you lowered three months earlier impresses nobody when DAUs (daily active users) come in at 138 million, down 9% and almost 8 million short of consensus. Hours fell 15%, monthly payers 16%. Q4 guidance: bookings down another 10 to 15%, negative adjusted EBITDA (profitability), negative free cash flow.
Kress's thesis is blunt. Unsurprisingly. The 2025 peak was built on Grow a Garden and the steal-from-other-kids genre; the founder and CEO Dave Baszucki doesn't want those games, and according to Eric, Roblox without them is a $35 stock, not a $138 one. The market seems to agree.
Personally, I think Eric is half right. Seasonally adjusted retention was the best since 2021; 18+ DAUs grew 18% and spend per payer rose 5%. Engagement didn't break, but conversion did. And fewer kids paying lines up with age checks and tighter safety rolled out earlier this year. Picking safety over a monetization genre is a legitimate trade. If they hadn't done it proactively, it could have been a death blow for the platform. I see it as short-term pain for a long-term gain. Too bad that public markets just don't price on a ten-year horizon.
Then there's AI, which Roblox reports as costs in dollars and benefits in activity. Kress called it the biggest spin he'd ever heard. Regardless, the activity is real. Over 80% of the top 100 games used AI tools, and Build (video), Roblox's prompt-to-game tool, has already 500,000 users and 70,000 published games in four small test markets ahead of its US launch in December.
In the meantime, two games made almost entirely with AI hit the top 20. The dollars are real too: infrastructure and safety spend up 24% to $258 million, with Q4 capex at $380 million. One of those AI showcase hits is called Break and Steal an Egg. Yes, another game relying on kids' emotional torture...
Supply has never been a constraint on Roblox. When a teenager in Serbia can prompt a game into existence over a weekend, the platform stops being a toolmaker and becomes an editor. Dave can tune discovery away from the games he dislikes. But he can't tune kids away from wanting them. And that's the dilemma.
Creator Ayo Omolewa decompiled Block Blast, Hungry Studio's puzzle juggernaut with around 70 million DAU and an estimated $127 million in ad revenue from January to May. The client pulls down hundreds of server-assigned JavaScript modules that govern difficulty and ads. One, which he identified as RewardHighValueAdHarvestTrait, waits until you've finished three rewarded ads, the loaded ad is worth 3.5 times the last one, and you're 1,000 points past your high score. Then it deals a hand you cannot place. The video passed 1.4 million likes on Instagram.
Laura pushed back on the podcast. Decompiled code shows half a system: server-side models are invisible, trait names drift from intent over years of tuning, and dynamic difficulty protects players as often as it squeezes them. Josh, who runs Wildcard Games, a primarily a monetized studio with a difficulty system he built, went further. Every ad earns money and raises churn. Rewarded video is the hard version of that trade, and Block Blast found the moment where a motivated player meets a high CPM. Josh summarized it perfectly: "Santa isn't real in gaming". Without that tuning, LTV (lifetime value) doesn't cover UA (user acquisition).
Laura and Josh are both right, and I think they both skip the part that matters. Engineering close losses is a craft. Keying a loss to the price of the ad rather than the state of the player is something else. When you tune the difficulty around fun, you're a designer. When you tune it around eCPM (effective cost per 1000 impressions), you're an auctioneer.
Furthermore, there's a detail most coverage missed: in Omolewa's own live test, the loaded ad priced at 1.94 times the last one, so the rule never fired. A rare edge case went viral as a death algorithm. Josh said outsiders were disgusted. Insiders asked how fast they could copy it.
The players who felt cheated weren't wrong, and the developers taking notes aren't villains. Free-to-play has always been a contract only one side could read. Decompiling handed the other side a copy. So here's the test before any trait ships: would you name this variable the same way if the player could see it? Because now they can.
Claude Code now declines to rebuild existing games. Image: Mashable
Last week someone spotted two new sentences in Claude Code's prompt for Opus 5.5. Build original games, genre games, and games from the user's own designs, but not a recreation of a specific existing game, "even if it is renamed or called original." Addy Osmani, formerly of Google Chrome and now at Anthropic, confirmed it, citing the usage policy against IP infringement. It follows weeks of people prompting models to rebuild GTA, Diablo II and Zelda.
Josh's reaction was... unprintable. And he's right on enforcement. Serious cloners run open-weight Chinese models on their own hardware, with no policy and no server bill. Teams already decompile every competitor update and map it against Sensor Tower revenue to see which code moved the needle. The window for profiting from innovation keeps shrinking as the cost of fast following falls. Kress got it right though, copying is worthless without a distribution power, and the exposed businesses are hypercasual factories.
Two sentences won't stop a determined cloners. And to be clear, they weren't written for cloners. They were written for IP holders, enterprise buyers, and courtrooms. This is simply a line that moves liability onto whoever crosses it. Real gatekeeping, if it arrives, will come from Apple review, Steam and Roblox discovery. Models are commodities, which is a whole another story. Storefronts still decide who gets into the store. But there's another cost that worries me more: studios are sitting on demos because a public build is now a blueprint.
But lets get real. code was never a moat. It was just expensive. What matters is an audience that trusts you, IP people care about, a team that ships the next update before the clone ships the last one. Following is not only immidiate but close to free now. If your advantage can be decompiled, it was never a moat.
Forza co-creator Dan Greenawalt is retiring after nearly 27 years. Image: PC Gamer
Two exits in one week. Dan Greenawalt, who founded Turn 10 and built Forza Motorsport, is retiring after nearly 27 years, with Turn 10 folded into Playground and Motorsport effectively parked behind Horizon. Helen Chiang, 21 years at Microsoft, former head of Minecraft and Xbox's first COO since July, left on October 23 with a one-line memo from Xbox CEO Asha Sharma. After reports of a policy violation, Chiang posted "Clarity is Kindness" on LinkedIn, denying leaks, joining a competitor, misconduct, and being passed over. What shed didn't say is why she left.
Kress's read was raw: the scarcest asset in a studio is the people who manage creatives and get games out the door, and losing them is how studios die. He called the inside absolute chaos. Jen made the operator's point. Rip the Band-Aid once, because a trickle of change keeps everyone braced. Meanwhile, Microsoft has offered voluntary exits to roughly 60,000 eligible employees, decisions are due in early December. The first hands up usually belong to veterans with options.
Folding Forza into Horizon is rational. I'd probably make the same call. The problem is sequencing: layoffs, studio mergers, a new COO, that COO gone in three months, then buyouts. Each move is defensible alone. Together they tell every senior person that tenure doesn't buy you protection. As a result, people stop taking creative risks and start taking recruiter calls.
Studios don't run on IP or engines. They run on a handful of people who matter. They are sometimes invisible on the org chart until they're gone. Every reorg I've seen promised focus. The ones that failed dragged on and were riddled with uncertainty. The best leaders I've worked with did the painful thing once, explained it, and let everyone get back to shipping.
โState of Game Advertising 2026 tracks how publishers in both the PC/Console and mobile space promoted their titles across 41 markets. You will not find more complete analysis on gaming marketing anywhere else!
Appcharge is your proven partner. $1 billion + in transactions processed across top titles. No platform middlemen. No fragmented handoffs between payment vendors. No watching your LTV models fall apart because you don't own the checkout experience.
Own your players. Own your revenue. Own your economics.
355. IP Won't Fix a Bad Game | Luis de la Camara, ex-Rovio, King & Sonic
Every studio says it wants a franchise. Luis de la Camara, who has led marketing on Angry Birds, Candy Crush and Sonic, says most can't define one. His test is simple: an IP only becomes a franchise once it has carried more than one product, and no IP will ever fix a bad game.
356. We Left Finland and Sweden for South of Europe. One Year Later
One year ago Phil left Sweden for Cyprus, and I left Finland for Athens. We compare the bureaucracy boss fights, how Cypriot and Greek taxes stack up against the Nordics, and the slow work of building a social life from scratch. If you've ever thought about moving your life and your studio south, start here.
Once operators, always operators. We're industry veterans who break down games, apps, and digital businesses for the people actually building them. Read by 15,000+ industry insiders weekly.
Hello friend, Every story this week sets conviction against evidence. Sands had the conviction to fund 440 people and no proof that a game would hold players. Brussels has conviction about harm to children and no data on the age split. Developers are convinced AI is hurting them and yes use it more than ever. The one company in this issue that runs on evidence keeps an unprecedented 84% margin. Conviction is what gets you started. Itโs the evidence that decides who is still standing when the...
TWIG #407: Roblox's Steal Economy, Tim Sweeney Goes All In on AI while Xbox Loses Its COO For maximum insight and entertainment, watch the full TWIG podcast with hosts Jen Donahoe, Eric Kress, Phillip Black and guest host Adam Telfer (YouTube / Apple / Spotify). Happy Friday, friend! Mishka here. I skipped this weekโs recording for a padel courts and dinners in Limassol and Tel Aviv. It's truly heartwarming to see you, our readers and listeners from around the world, in real life. Jen...
๐ Join us at our upcoming events ๐Executive Supper Club: Limassol - October 26The Padel Club: Limassol - October 27 Today's newsletter is brought to you by Sensor Tower and VYBS. More about them later... Happy Friday, friend! Mishka here. Every story this week is really about buying time. Shooters and sports games wait for the GTA wave to pass. Embark trades monthly content for systems that hold players between drops. Roblox bulls want to get out of quarterlies and go back to five years...