Pinned Tweet
After shipping viral single-purpose web tools, I wanted to build something with real depth. As a solo dev, I just launched 18 Days: Kurukshetra: a tactical Mahabharata card battler built with multiple AI agents.
Here is the stack, the tools, and what broke along the way: 🧵👇
1/ Why build this?
I have been gaming for 20 years. Thousands of hours across RPGs, turn-based strategy, and tactical card battlers like Gwent.
I also grew up with the Mahabharata. It is one of the richest, most dramatic war epics ever told. Endless lore, legendary warriors, tactical boons, vows, and divine astras (weapons).
Most importantly:
• Zero IP encumbrance (public domain)
• Nearly 1 billion people grew up with these characters
• Almost nobody has built a deep, tactical card game around it
2/ Respecting the lore: Hand-reviewed cards
I refused to build a game that just slaps Indian names onto generic fantasy cards. Every card, warrior trait, and ability was reviewed by hand against original texts. Examples:
• Bhishma cannot simply be destroyed in frontline combat while his death-vow stands.
• Karna fights without armor (having sacrificed his Kavach before the war); his gameplay revolves around his single-use divine dart Vasavi Shakti and fatal curse triggers.
• Divya Astras (divine weapons) are not random dice rolls. They are tactical spell strikes with exact requirements and counters.
3/ Two distinct ways to play
• Tactical Card Battler: 3-round competitive lane warfare with strict 170-point army budgets, where positioning and synergy decide the day.
• RPG Story Campaign (The Chronicles of Kurukshetra): A 60-chapter narrative RPG with branching dialogue choices, moral dilemmas, and tactical combat encounters.
4/ The big bet: Web-first
Everyone asks why I didn't launch on Steam or mobile app stores first. My reasoning:
• Zero friction: Anyone can tap a link and be in a battle within 3 seconds.
• Distribution control: No 30% app store tax, no arbitrary review delays, and no fear of sudden delisting.
• Rapid iteration: I can ship daily balance patches, UI polish, and story chapters instantly.
But fear not, Android and iOS apps are coming next :)
5/ How a solo dev built this: The agent setup
Building a game of this scale solo sounded impossible to me. The secret was running a dedicated trio of specialized AI agents on local branches:
• Claude: Core game engine reducer, rules, and mathematical balance.
• Codex: Test reliability, Playwright suites, and battle simulation checkers.
• Antigravity: UI presentation, card visuals, and story choreography.
Artwork and videos have been handled by workflows created in Adobe Firefly, Meta, Codex (GPT 2.5 family) and Gemini (Google's video-gen family).
6/ The upsets and breakthroughs
It was not a smooth straight line:
• Balance chaos: Early on, a single overpowered weapon broke 50% of matches. We had to build an automated simulation harness running over 10,000 bot battles overnight to keep card win-rates in check.
• Agent coordination: Managing multiple AI agents meant building strict merge gates so they didn't collide or overwrite work.
• Tactile feel: Making a web game feel like a native, responsive obsidian console on a mobile phone took weeks of CSS and animation refactoring.
7/ Follow the journey
18 Days: Kurukshetra is live right now. Free to play on any phone or desktop browser [link in comments]. I am going to build this in public and share everything:
• The exact tools and agent prompts I use
• Daily analytics and commercial traction
• Game balance math and marketing experiments
• Every upset and every breakthrough
If you are a builder, gamer, or indie hacker, follow me and come build this with me :)
udipta basumatari retweeted
embarked on the most exciting project of my life
it's not an agency
not a SaaS
not an app
it's .... 🥁
a whole domain extension
exactly like a .com or a .ai
far from the spotlight, yet a lifetime asset with strong economics
I raised $400k to apply to become the operator of ".labs"
and offer anyone a chance to own a [name].labs
the vision: the extension for builders and innovators
but wait... why the hell am I doing this? and how can we own an extension?
storytime:
almost 2 years ago, I launched yet another app
went on to buy a domain and the registrar says "invalid TLD"
the extension I typed simply didn't exist
I was specifically looking for a .labs, it sounded so nice and meant so much
but not a single one exists
being curious I looked up how domain extensions work and how new ones get created
turns out you can apply for one, but only when ICANN opens a round
the last round was in 2012, and the next one... 2026
14 years later
it stuck with me, it felt like the opportunity of a lifetime
there will only ever be one .labs
a name that promises progress
as a builder who believes in innovation
I knew I had found my project
the kind I could spend the next 20 years on
the application isn't cheap though: $227k just to get in
so I worked in stealth for almost 2 years and raised from top-class angels
took the leap, created a company, and sent the application
ICANN still has to review everything and big players want it too
but I’m going all in
on to building the largest community of innovators
if you want to be part of it in your own way
or want to waitlist your .labs name
jump in (link below)
I've already moved a lot of my work to Gemini models. They are fast, efficient and reliable.
My only gripe is that the models confuse me, I don't know when to use which.
Gemini 4 Neon: frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows.
Begins rolling out today in the Gemini API in @GoogleAIStudio, Gemini for business, for paid Google AI Pro and Ultra subscribers in Google @Antigravity, and for Google AI Pro and Ultra subscribers in the @GeminiApp.
11
Someone’s donation made Blender free for you. The next artist who opens Blender for the first time is counting on the same thing.
Now it’s your turn to invest in a brighter future: fund.blender.org
Today, we're asking for $10. That's it. If every active Blender user contributed $10 this month, the Blender Foundation would meet its entire funding goal for 2027.
#b3d #devfund
I've always been fascinated by the lifecycle of stars. Here's my videos about it made with Opus 5.5 and a custom harness
Gemini and Codex can do the funniest thing here and incorporate their Gen AI features to work alongside code.
[I've been doing this already but they still can't match the reasoning of Opus 5.5... yet]
This is most likely the latest Fable 5.5 checkpoint. The video is pure code.
Fable 5.5 will change the entire game. Im speechless. And yet, 99% of folks outside of our bubble have no idea.
h/t @chetaslua nitter.cf/chetaslua/status/21171…
1
48
Built this explainer video with Claude Code and multiple agents. No Generative AI here: every single frame is rendered directly from code.
How it works:
• Research agent drafts the script
• My voice recording
• Coding agent builds 3D scenes in three.js (headless Chromium)
• Whisper syncs the captions
• QA agent checks framing & overlaps
• Same code exports both 16:9 wide and 9:16 vertical
Stack: three.js, Playwright, FFmpeg, whisper.cpp, and Google Flow for the music.
Took 28 iterations to dial in. First attempt, but deterministic code gives you control video models just can't touch.
Will keep improving with future tries :)
udipta basumatari retweeted
Live gameplay from 18 Days. Round 3 decider: Pandavas vs Kauravas. ⚔️
Arjuna vs Karna on the board, and Divya Astras firing on both sides down to the wire. 🔥
Watch how the final turn unfolded. 🏹
Link in comments.
Balancing 3 asymmetric factions nearly broke my game. ⚔️
Solo Dev Log #4 is up:
How I ran 7,200 simulations to balance three lore-faithful Mahabharata armies, and why my tools kept lying to me.
Full breakdown below. 🛠️
#solodev #indiedev #indiegame
Free credits don't matter if the product is too slow to use. Codex keeps topping me up, but the performance is just impossibly lagging.
Shifted most of my work to Gemini on Antigravity and it runs like "Flash"
1
1
168
Solo Dev Log #3 - Video workflow
I already had the artwork. This time, I worked with Codex to turn it into a two-minute video using JS 🎬
Wrote up the process, the iterations and what still needed a human eye. Saved the workflow as a skill too. 🛠️
udipta basumatari retweeted
⚡ ANDROID CLOSED BETA IS LIVE ⚡
We need 15 ruthless testers to battle-test 18 Days: Kurukshetra 🏹
FIRST 15 TESTERS GET LIFETIME VIP:
- Free Asura Pass
- Full 62-Chapter Story Campaign
- No ads forever
Drop a comment below and we’ll DM you the link immediately! ⬇️
Rightly said, AI is just another tool. What matters is the output, and whether the audience enjoys the product.
When I was writing Unreal Engine 1, my time was divided as follows:
- Around 30% went to seriously deep coding on hard problems: assembly language texture mappers, C++ BSP trees, data persistence, object oriented systems design, and networking.
- Around 70% was just typing and refactoring large bodies of code needed for a complete engine feature set. This was not hard, and while I was doing it my mind was mostly on the deeper coding topics, thinking and planning for what's next. It didn't feel much different from vibe coding today, except that it was just me typing code on an IBM Model M keyboard rather than prompting it and reviewing the results.
So, it has always seemed perfectly logical that we would rely on AI for that 70%, once the AI could do it with a similar quality standard as we could, which has never been perfection. It came as no surprise that, after a few years of AI model evolution, it was able to take the reigns on that 70% and free up programmers' time for more valuable tasks.
What's more of a surprise is the emergence of AI as a sort of insta-Carmack able to solve hard technical problems quite cleverly and often cleanly, and that it's able to reason about and work in complex 3D scenes both by operating the tools and by directly understanding the raw data files.
I abandoned assembly language in 1998 when out-of-order CPUs made much simpler C/C++ code faster, and never I missed it. If prompting through architectural documents, concept art, and design plans can transcend typing code and tweaking vertices, that will enable developers to spend their time on higher-level, higher value technical and creative tasks.
What matters is the end product, whether it's a game or an engine. Anything developers can do to try out more ideas and make better software and games faster is a win for everyone.
















