SuperbulletAI launched the most powerful AI Game Builder for Roblox, and it's free for everyone to try

Resources Community Resources
114

this is cool but let us keep our jobs.

115

obv lol, this AI can’t work without u :stuck_out_tongue_winking_eye:
just makes you much more faster and smarter

maybe for javascript/typescript, but not Roblox Studio, which uses Lua. it’s pretty easy to hyperfocus an AI which will theoretically be 100x better if you use the right method. Even if it is just one person.

BulletLearnV1-Beta, which is 50x cheaper, allows you to learn and build faster. In contrast, Claude Sonnet 4.0 will probably take $4,000 and 1 month to build one due to its inference speed and token consumption

1 Like
116

I pretty sure the only one being smart is the AI. A bunch of noobies will use this to make a quick cash-grab game :confused:

1 Like
117

Please tell me your website isn’t built by builder.io …

118

So, does this is just cursor and claude but for roblox games?

119

Launched the following models for v0.2.04 enabling free users to access it and it will feel unlimited:

1. BulletMindV1-Beta

  • About 1.2× faster than Claude or GPT-5.
  • Costs about 8× less than Claude Sonnet 4.0.
  • Almost the same quality as Claude or GPT-5.

2. BulletMindV1-Fast-Beta

  • Same model, but insanely quick—about 27× faster.
  • 2.5× cheaper than Claude Sonnet 4.0.
  • Still just slightly below Claude/GPT-5 in quality.

3. BulletLearnV1-Beta

  • 50× cheaper than Claude Sonnet 4.0.
  • 3× faster than Claude/GPT-5.
  • Smaller model, but still surprisingly powerful.

4. BulletLearnV1-Prime-Beta

  • 10× cheaper than Claude Sonnet 4.0.
  • 54× faster (though this one might cost a bit more to run).
  • Around 1.6× stronger than BulletLearnV1-Beta.
120

It’s a platform for building Roblox games, designed to be conveniently easy for developers to get a game of their own started. It’s hyperfocused on Roblox, yes.

Sorry no, I wish..

121

i think OP just wants to flex how hes awesome, he’s been on roblox for 9 years, he has 200m visits on his games, built games with billions of visits total, bought some stupid korblox package, wants to support the Inevitable Robot Uprising That Will Kill Us All, and nobody can tell me otherwise

122

Gave it a try, since OP made it might as well try it for fun.

The daily freebie tokens seems to evaporate within 3 mins with a basic prompt and the daily reset doesn’t seems… to be functionnal?

123

There’s a problem with BulletLearnV1 so we temporarily redirected it to our other models leading to expensive token costs.. We’ll try to have it fixed today.

124

0% better, his BulletMindV1-Beta is just Qwen3 coder served by OpenRouter

125

In its earlier iteration, we tested many models, yes. Also includes DeepSeek, Qwen, Kimi, Meta - ever since then, we’ve migrated to a dedicated infra/server partner to run our AI model. This is why it has “-Beta” tag in it, the official release of the “-Classic” is in months from now but:

  • We are releasing the first benchmarking suite for Roblox game development soon and even GPT-5 won’t be able to compete against our AIs.
  • Imagine centralized car system with 50 cars. All imported in your game in just 15 minutes. Which can be done in 15 minutes vs 2 months
126

From my experience, this is not a very high bar. GPT-5 feels like it was made to cut costs, not to be a better model.

2 Likes
127

Valid take, but I’m including all the SOTA models in the benchmarking results like Claude Opus 4.1 which costs prolly 90x more than our AI model lol

128

You using the internet and playing Roblox games is contributing to climate change. So is using your overpriced device that contributed to climate change to build and continue to use. This logic makes no sense.

3 Likes
129

Although I’m not a fan of AI technology being used in a way such as this one, this argument just makes you come off as completely ignorant and that you don’t know what you are talking about. For reference on how exactly AI tech can actually be far worse for climate change than conventional computing (AI technology is many times more resource intensive than conventional computing) I made a post earlier in this thread linking to some articles on the subject.

2 Likes
130

Claims that AI data centers will worsen drought are overstated, since companies are already deploying zero-water cooling and shifting toward renewable energy.

https://www.reuters.com/legal/legalindustry/strategies-promote-sustainability-data-center-buildout--pracin-2025-08-28/

or techradar’s article:

Furthermore, general internet services such as social media, netflix, or even your searches in Google takes more resources than AI.

“At present, Goldman Sachs Research estimates the power usage by the global data center market to be around 55 gigawatts (GW). This is comprised of cloud computing workloads (54%) (netflix, youtube, facebook, twitch), traditional workloads for typical business functions such as email or storage (32%) (gmail, your hard drive and more), and AI (14%).” > Source: Goldman Sachs Research.

According to OECD, they projected AI operations to account only for about 0.5% - 0.7% of the US’s total annual water withdrawals by 2027.

AI enhances advancement in technology even for climate change innovation, and it may ultimately be the very force that saves humanity from our own overpopulation and the looming exhaustion of vital resources like water.

1 Like
131

In my opinion we shouldn’t be trying to make something that’s basically doing everything for us.

Let the people learn by themselves, The forums is here and the sources are there.

2 Likes
132

When you factor in that said industries are countless times larger than AI currently is, the argument and comparison that they take up a lot more resources is completely meaningless. Consider how small the AI industry is currently in comparison to these other ones and suddenly that mere “14%” looks a lot more concerning since that’s a lot more usage for technology that’s not even remotely as widely used as conventional computing is.

1 Like
133

I am talking about “electricity” here, and only for servers or data centers.

According to the International Energy Agency reports that in 2024, data centers consumed around 415 TWh, which equals roughly 1.5% of total global electricity.

That means AI’s “14% of data center use” translates to ~0.2–0.3% of world electricity, tiny compared to the ~25% used by industry, ~20% by residential, or ~40% by agriculture + manufacturing combined. FOR ELECTRICITY.

You were talking about cost efficiency of “water”. Electricity is a different subtopic. My point stands.

AI is an INDUSTRY itself and the adoption to all companies are happening now, yet it only consumes less than what the whole world are already consuming.