Tech Debt Club

Valentin and Bobsi met contracting for the same multiplayer games platform, where a switch to Photon Quantum brought solid determinism and a workflow that took two hours to network a rolling ball. Almost two years after they started building PurrNet out of that frustration, their MIT-licensed Unity networking solution has passed 3,000 Discord members and around 20,000 GitHub clones a month. We talked about retrofitting multiplayer into ten-year-old single-player code, why no tutorial teaches server authority, a 100-player Unity server using 100–150 MB of RAM, player IDs that survive reconnects, 550 automated regression tests, and shipping CRACKED and Arctic Drive on their own stack. Every networking solution can make games, Bobsi says — two percent less bandwidth was never the point.

🔹 Links:

PurrNet on GitHub: https://github.com/PurrNet/PurrNet

Bobsi Tutorials on YouTube: https://www.youtube.com/@BobsiTutorial

NowUI: https://github.com/BlenMiner/NowUI

CRACKED on Steam: https://store.steampowered.com/app/3917010/CRACKED/

Arctic Drive on Steam: https://store.steampowered.com/app/4009490/Arctic_Drive/

🔹 Chapters:

• 00:00 Intro
• 01:18 What We've Been Playing
• 19:34 Voxels, C++ in Unity, and a UI Library Born From Hating UI
• 25:41 Ten-Year-Old Code Bases and Retrofitting Multiplayer
• 36:16 Arctic Drive, and Yes, You Can Pee on the Snow
• 41:33 Two Hours to Network a Rolling Ball in Photon Quantum
• 48:57 Server Authority, the Thing No Tutorial Teaches
• 53:48 A 100-Player Unity Server in 150 MB of RAM
• 57:51 Two Percent Less Bandwidth Doesn't Matter, Main Character Syndrome Does
• 01:02:30 Player IDs That Survive Reconnects, and Swapping In Your Own Systems
• 01:06:47 Dogfooding, Discord Growth, and Traffic From ChatGPT
• 01:15:50 550 Automated Tests and 45,000 Playtesters
• 01:20:45 CRACKED, Rollback Prediction, and Falling Sand With No State
• 01:24:55 One Book, One Movie, One Game, and WinRAR vs. OpenAI

Episode URL: https://tech-debt.club/episodes/9-tdc-09-from-photon-quantum-to-purrnet-with-valentin-and-bobsi

Rytis Jadzevičius came to games in 2019 from government work, commercials and digital marketing — not a programmer, but always surrounded by them. Seven years later his studio, Fluxo Games, has ten people, a submarine game still in early access, and Book Smugglers, a self-published game about Lithuania's 19th-century press ban that took three on-and-off years to finish. We talk about surviving on one in-house and one client project at a time, what happens when government and work-for-hire money dry up together, emailing 10,000–20,000 influencers after parting ways with a publisher, whether this downturn is AI or post-COVID overhiring, why a country of three million needs its own game awards on national TV, and porting to Switch with a dev kit you can no longer get. His lesson for everyone: kill more games before you release them.

🔹 Links:

Fluxo Games: https://fluxogames.com/

Book Smugglers on Steam: https://store.steampowered.com/app/2720940/Book_Smugglers/

Rytis on LinkedIn: https://www.linkedin.com/in/rytis-jadzevi%C4%8Dius-5696a451/

ASML EUV lithography: https://www.asml.com/en/products/euv-lithography-systems

🔹 Chapters:

• 00:00 Intro
• 01:29 What We've Been Playing
• 09:23 Not a Programmer, but Always Surrounded by Them
• 13:06 One In-House Game, One Client Game, and Why Not Mobile
• 21:57 Bigger Than the Post-COVID Crash, and Don't Blame AI for Overhiring
• 28:18 GTA 6, Crysis, and Why Hardware Never Gets Cheaper
• 37:55 Plasticalypse and "We Should Have Killed More Games"
• 45:48 Work-for-Hire Dries Up: Pivoting to QA, Porting, and Influencers
• 50:31 Book Smugglers: A Banned Language, a Lost Publisher, and Breaking Even
• 53:48 Free Marketing, Raid Shadow Legends, and Communities Paid in Coins
• 01:01:25 Playing the Patriotism Card (and Kevin Bacon)
• 01:09:31 Most Studios Would Have Killed It Three Times, and a Switch 1 Dev Kit
• 01:14:18 A Cooking App in the Game Accelerator, and What Poland Got Right
• 01:19:32 A Statistics Code for Game Dev, and Game Awards on National TV
• 01:29:12 Rytis's Advice: Think Twice, the Plates Have Shifted
• 01:31:04 One Book, One Movie, One Game, and Warcraft III Custom Maps

Episode URL: https://tech-debt.club/episodes/9-tdc-09-making-games-in-lithuania-with-rytis-jadzevicius

Mike Geig joined Unity thirteen years ago as a screencaster and is now a principal developer advocate — a job with no sales quota, no marketing targets, and a single KPI: how happy the community is. He'll have flown close to 100,000 miles this year to find out, and he calls himself a game developer first and a Unity employee second. So we brought him the arguments Unity developers actually have: Godot overtaking Unity at the GMTK Jam after nine years, an engine that only started paying for itself in the last two years, the open-source fight he's been losing internally, the DOTS animation system that got cut, whether HDRP is really dying, and whether Unity AI can compete with a $200 Claude subscription. He'll never lie to us, he says — but he reserves the right to be wrong.

🔹 Links:

Mike on X: https://x.com/mikegeig

Mike on LinkedIn: https://www.linkedin.com/in/mikegeig/

GMTK Game Jam 2026 engine stats: https://bsky.app/profile/gamemakerstoolkit.com/post/3mrmx5mi3mk24

Yuri's Unity AI video: https://youtu.be/JsYYnVemozo

After Effects MCP: https://github.com/Engine-Room-Games/after-effects-mcp

🔹 Chapters:

00:00 Intro

02:23 What We've Been Playing

10:44 100,000 Miles and a KPI Called Happiness

24:18 Game Developer First, Unity Employee Second

31:15 Two of Everything, Addressables, and Why Nothing Is a Two-Week Fix

40:52 Engine Upgrades and the Lies We Tell Ourselves

47:28 Godot Dethroned Unity (at Game Jams, Anyway)

54:03 The Open Source Fight Mike Has Been Losing for Years

01:01:06 The Engine Only Just Started Paying for Itself

01:06:06 Digital Twins and Unity Studio: The Non-Game Side

01:19:07 The Hundreds of Games You Didn't Know Unity Made

01:31:43 DOTS Has No Animation System, So Don't Make a DOTS Game

01:40:37 Thousands of Empty GameObjects, GetComponent, and Git Blame

01:48:33 HDRP Isn't Dying, URP Is Becoming Universal

01:55:26 Unity AI vs. a $200 Claude Subscription

02:05:59 AI Witch Hunts, and Why Yuri Didn't Go Indie

02:15:30 Mike's Unity Tips

02:18:41 One Book, One Movie, One Game, and Pokémon Go

Episode URL: https://tech-debt.club/episodes/8-tdc-08-13-years-inside-unity-with-mike-geig

Daniel Mironov left Google to help start vgames in March 2020 — a games-only VC fund in a country with two dozen cybersecurity funds and none for games — roughly a month before the world locked down and mobile went vertical. Six years and nearly 50 investments later (SuperPlay and InnPlay Labs both went to Playtika, Candivore is still printing money with Match Masters), he's a partner at the fund and has just launched a $10M Indie Fund that hands PC and console developers project financing without taking a single frame of their IP.

🔹 Links:
vgames: https://www.vgames.vc

Indie Fund announcement: https://gamesbeat.com/vgames-raises-10m-fund-for-project-financing-for-indie-games-exclusive-interview/

🔹 Chapters:

00:00 Intro
01:46 What We've Been Playing
09:57 vgames, or How to Start a Games Fund in March 2020
20:36 One in Fifty, and Why the Job Is Saying No
33:38 Show Me a Prototype, Not a Deck
47:28 What Turkish Founders Are Better At
53:05 What $3–4 Million Actually Buys
01:12:34 China, IP Deals, and the Platform That Never Came
01:20:17 The Palworld Question
01:29:36 The Indie Fund: Capital Without a Publisher
01:51:32 AI on Steam and the Jonathan Blow Exception

Ivan Murzak spent the early 2010s shipping Evil Cogs, a Limbo-style mobile platformer that hit 5 million downloads on Android by squeezing performance out of phones that had no business running it. A decade later, he's a Microsoft MVP maintaining what's become the most-downloaded open-source Unity MCP server - the one anyone trying to get Claude, Cursor, or Copilot to actually do useful work inside the Unity editor probably ends up using.

🔹 Links:

Unity MCP: https://github.com/IvanMurzak/Unity-MCP

Ivan's YouTube: https://www.youtube.com/@Ivan_Murzak_Dev

🔹 Chapters:

• 00:00 Intro
• 03:22 What We've Been Playing
• 13:49 Evil Cogs: 5M Downloads and Limbo by Accident
• 44:09 Why You Shouldn't Give Your Game to Your Mom
• 01:06:17 Building the Most-Downloaded Unity MCP
• 01:20:05 The MCP Zoo and Why CLI Is Winning
• 01:48:46 Reflection All the Way Down
• 02:04:11 Does the CS Degree Still Make Sense?

Episode URL: https://tech-debt.club/episodes/6-tdc-06-from-indie-game-dev-to-top-unity-mcp-with-ivan-murzak

Andreas Gielov spent three years helping run Brackeys, one of the most influential Unity YouTube channels of all time. After Brackeys stopped producing Unity content, he went on to manage a string of game development YouTubers before joining Sirenix — the team behind Odin Inspector and the upcoming UI library Pangui — as VP of Marketing.

We talk about what actually makes a game dev YouTube channel work, why advocate-driven marketing beats paid ads in niche markets, and why Sirenix has been quietly building Pangui for five years before letting anyone see it. Along the way: the tech stack is bloated, your in-house Unity UI framework probably shouldn't exist, and Jonathan Blow gets brought up (again).

Chapters

• 00:00 Inside Brackeys and the Reality of Game Dev YouTube
• 48:18 Buying 70 Odin Licenses (And What Happened Next)
• 54:23 Why Odin Worked, and What Pangui Actually Is
• 01:02:51 Five Years of Silence on Pangui
• 01:08:21 Rebuilding UI From Zero
• 01:14:36 Open Source, AI, and Why You Shouldn't Build Your Own UI

Episode URL: https://tech-debt.club/episodes/5-tdc-05-from-brackeys-to-pangui-with-andreas-gielov

Lior Hadashian, co-founder and CTO of Gavra Games, joins Yuri and Amit to talk about building and shipping Warriors Rise to Glory — a turn-based gladiator fighting game inspired by a goofy Flash game from the internet's ugly years. What followed was five years of bootstrap chaos: a Bilibili influencer blowing up the game overnight, a Chinese localization done in one week, a main font that turned out to be stolen, two appearances on Israeli Shark Tank, and a mobile pivot that never happened.

They also go deep on the realities of indie investment — what it actually means to raise money, why turning down $130k on live television was the right call, and what "alive but not kicking" looks like for a studio winding down.

The back half gets technical: Unity version upgrades and why they break in exactly three places every time, Google's ongoing war with Unity Package Manager, and Amit's experience upgrading a 4-year production project to Unity 6.3.

Lior is working on something new. He won't say what it is. He's on Unity 6.4.

Topics covered:

• Warriors Rise to Glory — the game, the inspiration, the Venn diagram problem of turn-based fighting games
• Launching on Steam without localization infrastructure, then scrambling to add 10 languages
• How a BiliBili influencer caused a Chinese localization crisis in week one of early access
• The font that was illegal (and why you can't just pay for it)
• Israeli Shark Tank — declining the first deal, coming back with receipts, and closing the second
• What $40k of initial investment and $257k in gross revenue actually gets you
• Why going mobile fell apart (the chicken-and-egg problem of mobile talent and capital)
• Unity upgrade pain: asset bundles, TextMesh Pro, shaders, and Google's UPM situation
• Swap Heroes — Amit's game, go download it, leave a review (constructive preferred)
• "Your brain is the product"

Time Codes:

• 00:17 – Cold open: what are you playing
• 08:32 – Meet Lior / Warriors Rise to Glory
• 14:33 – The Venn diagram problem
• 15:13 – Localization: the full saga (BiliBili, 20% refunds, one week, illegal font)
• 37:36 – Shark Tank: twice, one rejection, one deal
• 47:52 – The numbers, the aftermath, the pizza
• 1:00:00 – Mobile, burnout, and the end of Gavra Games
• 1:05:37 – Unity upgrades: what always breaks
• 1:15:23 – What's next

Guest: Lior Hadashian

Episode URL: https://tech-debt.club/episodes/4-tdc-04-from-shark-tank-to-steam-with-lior-hadashian

Karlo Kilayko has been making games since before most of us knew what a game engine was - literally. From programming a CD-ROM murder mystery in C with one reference book and no internet, to shipping 30 mobile games a year across 382 devices at THQ, to becoming one of Unity's earliest professional users, Karlo has lived through just about every era of this industry.

We talk about what it was actually like to break into game dev in the 1980s, the brutal carrier-controlled mobile landscape before the iPhone rewrote the rules, and why the Nokia N-Gage might be worth more on eBay than you'd think. We also get into the conversation that doesn't go away: AI, what it can actually do for developers today, and the question nobody seems to have an answer for - where do juniors gain the foundational experience they need to use it effectively?

The second half turns toward shipping - or more honestly, the reasons people don't. The core philosophy, borrowed from a Trip Hawkins one-liner: you make zero billion dollars on games you don't ship.

We also cover object-oriented vs. data-oriented design, the game engine vs. game IDE distinction, clean code debates that get surprisingly personal, and how Unity's MCP limitations are affecting AI workflows right now.

Chapters:

• 00:00 - Introductions
• 00:32 - What we're playing
• 05:33 - Why people make things
• 09:35 - Breaking into game dev in the 80s
• 13:00 - Programming a CD-ROM murder mystery
• 17:15 - Data-oriented design, then and now
• 20:30 - ECS, composition, and Raylib
• 22:00 - Being an early Unity user
• 26:00 - Mobile games as business applications
• 31:00 - The Unity runtime fee debacle
• 34:00 - AI engines and the engine vs. IDE distinction
• 38:00 - Unity, MCP, and AI workflows
• 40:00 - AI amplifies what you already have
• 43:00 - Clean code debate
• 51:00 - Vim, terminals, and knowing the basics
• 55:00 - COBOL and legacy job security
• 1:03:00 - When designers use Cursor directly
• 1:05:00 - THQ Wireless: 30 games, 382 devices
• 1:09:00 - Nokia N-Gage and early mobile multiplayer
• 1:13:00 - Creative QA: elevators and microwaves
• 1:16:00 - Carriers, control, and the iPhone
• 1:19:00 - Better done than perfect
• 1:27:00 - The 90/90 rule and shipping frameworks
• 1:33:00 - Wrap-up

Guest:

Karlo Kilayko - game developer and producer.

Episode URL: https://tech-debt.club/episodes/3-tdc-3-40-years-of-shipping-games-with-karlo-kilayko

In this episode, Yuri Sokolov and Amit Netanel talk about what AI is actually changing in software and game development - and what it is not. They dig into the difference between productive and destructive AI workflows, why experience still matters when using LLMs, and why most developers and companies still struggle with architecture, tooling, and learning the right lessons at the right time.

They also get into Unity’s recent AI announcements, MCP servers, world models like Genie 3, startup culture versus corporate bureaucracy, the culture around bad practices in gamedev, and how social media shapes shallow technical opinions. Toward the end, they circle back to Bethesda-style engine hacks, AI hype in the industry, and whether AI is really replacing developers or just reshaping the work.

Chapters:

• 00:00 - Will AI take our jobs?
• 00:15 - Waiting on AI
• 01:45 - Good vs bad AI workflows
• 04:02 - AI for real dev work
• 06:13 - CAD and deterministic thinking
• 09:54 - Players want fun
• 12:46 - DI vs singletons
• 14:13 - Learning architecture through pain
• 18:56 - Startup vs enterprise culture
• 21:00 - Growing company pains
• 23:48 - Low barrier engineering culture
• 30:11 - LinkedIn brain rot
• 32:13 - Bad gamedev takes
• 36:49 - Clickbait and shallow opinions
• 39:32 - Focus and empty calories
• 44:00 - Unity AI announcements
• 45:26 - Genie 3 and AI hype
• 50:19 - Unity MCP
• 55:31 - AI for investors
• 58:13 - Where AI actually helps
• 1:00:25 - Extreme AI game jam
• 1:02:51 - AI amplifies skill
• 1:05:29 - Prompting with constraints
• 1:08:43 - Debugging with AI
• 1:10:29 - Fallout 3 train hat
• 1:12:55 - Bethesda engine hacks
• 1:15:18 - Building while AI works
• 1:17:59 - AI changes the work
• 1:19:01 - Is programming dead?
• 1:21:51 - Build vs buy
• 1:23:31 - SaaS needs real value
• 1:26:27 - Losing knowledge to AI
• 1:28:07 - Seniors using AI badly
• 1:30:24 - Could AI replace devs?
• 1:31:50 - Layoffs and AI optics
• 1:33:35 - Unity promises
• 1:35:47 - Wrap-up

Episode URL: https://tech-debt.club/episodes/2-tdc-2-has-ai-killed-game-dev-or-are-you-just-lazy

The conversation begins with an introduction to podcasting and transitions into a discussion on gacha game development challenges. It then explores the influence of existing games on new game development, highlighting the impact of popular games on the development of new titles. The conversation delves into the world of game development, exploring topics such as linear storytelling, user experience, technical challenges, data analytics, player decision tracking, preservation of classic games, and parental controls. The speakers share insights on game design practices, optimization, and the impact of player feedback on game development. They also discuss the importance of preserving classic games and the challenges of parental controls in regulating child access to games.

Takeaways

• Gacha game development challenges
• Influence of existing games on new game development
• Game development practices
• User experience in game design

Chapters

• 00:00 Influence of Existing Games on New Game Development
• 38:45 Game Design and Linear Storytelling
• 45:28 Optimization and Technical Challenges in Game Development
• 51:27 Player Decision Tracking and Game Endings
• 01:05:37 Parental Controls and Child Access to Games

Episode URL: https://tech-debt.club/episodes/1-tdc-1-making-games-in-the-shadow-of-other-games