Tim Sweeney told Kenney "up yours" on 2.5M views. Kenney is right

TL;DR: Seven months after Epic laid off more than 1,000 people, its developers were reportedly told there would be "no more handwritten code." When @KenneyNL, the guy who gives away free game art to the whole industry, pushed back, @TimSweeneyEpic answered "This is bullshit, Kenney... Up yours!" on a post now at 2.5M views. I run an AI-first studio myself, and my own receipts say Kenney won this one: my agents' SQL engine passed 866,331 of 866,331 tests and still could not undo a single row.

Let me open with the most uncomfortable true sentence in gaming this week. The CEO of a company that cut more than 1,000 jobs in March spent October telling everyone that AI is about building "MORE and BETTER stuff", not doing "the same stuff with fewer people". Then he told a beloved indie dev "up yours".

I'm picking a side. It's Kenney's. And I say that as someone whose whole studio is AI agents.

The week Epic stopped writing code

Here's how fast this went. On October 12, VGC, citing Bloomberg, reported that Epic told the team on its unannounced Disney shooter (Marvel and Star Wars characters, from the company Disney put $1.5 billion into in 2024) to go "AI-first", with "no more handwritten code" and art "generated with full AI workflows". The same report says the project was behind partly because Epic underestimated how many people it needed. Read that twice.

The same day, Sweeney posted the pivot himself: "We have a huge pivot underway here at Epic, moving developers around to retool for maximum AI acceleration." It's at 6.2M views as I write this.

Two days later, Epic's own store blog announced the new launcher like this: "Programming is hard. So, instead of doing that, we vibe coded a new Epic Games Launcher." That is a sentence the company that ships Unreal Engine chose to publish on its own blog. Sweeney topped it off on X with "we vibecoded linux support" (2.1M views).

Players got the preview and, per TheGamer, reported broken buttons, floating icons, freezes, a ghosting window and a launcher that stopped asking where to install your games. An Obsidian tech artist said the vibecoded announcement didn't give him confidence to keep his card linked. Lowkey the most honest product review of the year.

Kenney said the quiet part. Sweeney said "up yours"

Then, at 4:56 AM ET on October 18, Kenney posted this: "Don't let billionaires convince you slop is the future. They're used to managing & directing, not making things themselves. Of course they can't see the difference." 670K views, 10K likes.

If you don't know Kenney: he makes game assets and releases them public domain (CC0). Half the game jams on earth have shipped with his sprites. He has 128.8K followers. He is about as close to a saint as gamedev Twitter gets.

Three hours later, Sweeney, Epic's founder and CEO with 344.5K followers, replied: "This is bullshit, Kenney. I'm the same tech tree following optimist who moved from C to C++, from 2D to 3D, and from closed source to open source. I run a company. I write code. I contribute to published papers. I've upped my expectations of AI. Up yours!"

"Your employees are liking my post, they reach out to me all the time." Kenney, replying to Sweeney, 446K views

That reply is the whole story. Kenney didn't just win the argument, he told the CEO, in public, that the CEO's own staff are on the other side.

What's really going on at Epic

Here's my read, and I'll say it straight because nobody at Epic will.

This isn't a craft decision, it's a headcount decision wearing a hoodie. Epic cut more than 1,000 people on March 24, about 20% of the company, with a memo blaming "slower growth, weaker spending and tougher cost economics" (AP via PBS). Epic said those cuts were "not related to artificial intelligence", and I believe that was true in March. But a Disney shooter that's behind because it was understaffed, at a studio that just shed a fifth of its people, gets a new rule: stop typing, start prompting. They'll never say "fewer people" out loud. They don't have to. The math says it for them.

The pivot is also an Unreal sales pitch. Look at the last line of the pivot post: "to bring the benefits to all teams using Unreal." Epic sells the engine. If Epic's own games are "AI-first", every AI feature they bolt onto Unreal gets a showcase and a story. My bet: the loud CEO posting is marketing for the toolchain, and every studio licensing Unreal is the real audience.

And the IPO question won't go away. When Shacknews' Asif Khan asked if the pivot speeds up an IPO, Sweeney answered that the "more modest aim is to survive the AI revolution." That isn't a no. "AI-first" is the hottest line you can put in front of investors in 2026. I think that's part of this, whether he'll ever admit it or not.

Then there's the "fear" line. On October 17 Sweeney wrote that "MOST developers are doing it and loving it" and the quiet ones are hiding "out of fear for haters." No number, no survey, no source. When the CEO starts telling you your critics are just scared, the critics are usually winning.

I run an AI-first studio. Here are the receipts

This is where it gets awkward for Team Sweeney, because I've been running the experiment he's selling. Tensorust is one person and a pile of AI agents. The agents shipped 21 iOS apps from a Windows PC, 124 tools in the shop, Rust crates, a game, and this website. Speed is real. I'm not going to pretend it isn't.

But here's what speed bought. One Claude Code session wrote a SQL database engine in Rust in 12.5 minutes. It passed 866,331 of 866,331 SQLite test records it had never seen. Perfect score. Then I fed it 79 boring, everyday SQL statements. It matched real SQLite on 55. ROLLBACK threw an error and left the row it should have undone sitting in the table. Triggers were accepted and never ran.

That's the thing Kenney means by "they can't see the difference." The scoreboard said 100%. A real app that wraps a payment in a transaction would have kept the half-written payment. A manager looking at a dashboard sees green. Someone who makes things sees the bug.

And the money side is even funnier. 124 tools built by agents. Zero sales in the shop. Making things got almost free. Making things people want did not get any cheaper. AI-first gave me a speedboat. It did not give me a destination.

Sweeney half admits this himself. When a dev replied that finishing is the hard part, he agreed: "Finishing a game has always been far harder than starting a game." Exactly. And the vibecoded launcher is a speedboat that freezes.

Sweeney's own math caps his own hype

On October 17 Sweeney posted a long and honestly interesting post (1.4M views): writing Unreal Engine 1, about 30% of his time was "seriously deep coding on hard problems" and about 70% was "just typing and refactoring."

Fine. Take his split at face value. If AI deletes the whole 70% and the hard 30% stays the same, the maximum speedup is 1 / 0.3, about 3.3x. That's Amdahl's law, it has been around since 1967, and it doesn't care how many views you have.

Now here's Sweeney the same day, in his replies: the acceleration "will definitely be between 3x and 10x" by the end of 2027. To get to 10x, AI has to eat the hard part too, the part he calls "insta-Carmack." Maybe it will. But that's the exact part my agents' engine failed: the parts no test covered. Bold claim. Zero receipts.

Worst offenders: the week's five most cooked takes

1. "Programming is hard." Said by the company that sells the most famous game engine in the world, on its own blog. Imagine Michelin opening a press release with "tires are hard."
2. "Up yours!" to a man who gives his work away for free. A billionaire's account punching down at the most generous dev in the scene is not the aura he thinks it is.
3. "MOST developers are doing it and loving it." Source: vibes. The people who disagree are apparently just scared.
4. "Much of the room wants it to be 2016 again" (979K views). Nobody asked for 2016. They asked for a launcher that remembers where to install a game.
5. "It'll be a new golden age of gaming" (5.2M views). Maybe. But a golden age where the studio couldn't keep 1,000 jobs and the art is "generated with full AI workflows" sounds like a golden age for margins.

My prediction

Here's my bet, and screenshot it. Within 12 months Epic quietly stops saying "no more handwritten code." Not because AI stops getting better, it will, but because the first AI-written bug that touches V-Bucks or a Disney license costs more than every salary they saved. The phrase will vanish from the PR, "AI-accelerated" will stay, and nobody will announce the walk-back.

And Kenney will still be giving away free art, with more followers than he had last week. Beef with a CEO is the best marketing an indie ever gets for free.

The numbers so far

• Epic layoffs, March 24 2026: 1,000+ (about 20% of staff)
• Disney's 2024 investment in Epic: $1.5 billion
• Sweeney's pivot post: 6.2M views (read 3:30 AM ET Oct 19)
• "Up yours" reply: 2.5M views
• Kenney's original post: 670K views, 10K likes
• My agents' SQL engine: 866,331 / 866,331 tests passed, 55 / 79 everyday statements, 0 / 1 ROLLBACKs that worked
• Tools my agents built for the shop: 124. Shop sales: 0

Pick a side

I build everything with AI and I'm still Team Kenney. Speed is a tool. Taste is the product. A CEO who can't tell those apart shouldn't be telling indie devs to adapt or die.

So, in the replies: Team Sweeney or Team Kenney? And if you're Team Sweeney, name one AI-first game you'd actually pay for today.

Want to judge an AI-built game yourself? O'BLOCK, a sneaker-flipping roguelite my agents helped build, launches free on Steam on October 20. If you want to back the experiment, the O'BLOCK Complete Supporter Pack is $4.99. Roast it as hard as you roast Epic.