Project summary

aicraftcyberforge.com is a live, gamified educational simulator designed to train the next generation of AI software engineers on defensive prompt engineering, runtime evaluation, and structural safety primitives. Rather than teaching abstract concepts, the platform drops developers into concrete sandbox environments where they must defend applications against live security vulnerabilities. The curriculum actively challenges students to harden LLM integrations against adversarial attacks, isolate unsafe execution loops, and build robust alignment filters from the ground up.

What are this project's goals? How will you achieve them?

The rapid expansion of multi-agent architectures has severely outpaced developer safety training. Most software engineers understand standard application security but treat LLM boundaries as black boxes, making production apps highly vulnerable to prompt injections and boundary escapes. Cyber Forge solves this human-layer bottleneck by making AI safety practical, accessible, and deeply interactive.

By scaling this platform from a direct B2C educational tool into a comprehensive institutional resource, we aim to establish a verified baseline of secure coding practices for developers globally.

How will this funding be used?

I've set two different benchmarks for funding goals and uses.

the first being the operational floor level being 2,500 - 3,000 this would go toward zero cost learner access, by subsidizing sever less edge functions and proxy bandwidth for 10,000 guest sessions. a 10 level curriculum expansion specifically targeting emerging vulnerabilities. and university and ai safety students groups outreach and making packable turnkey workshop materials.

The second benchmark for funding is 15,000 - 25,000 for a full sprint on development including checkpoints like open community adversarial arena: to build a public sandbox where researchers and developers can submit custom defensive and let automated adversarial stress test challenge them in real time. open source evaluation harness: document and open source the automated headless evaluation runner so other research labs can benchmark defensive prompts locally. then red teaming tournament: this would be designed to offer cash prizes for jailbreak discoveries and defensive architectures. also serving as a advertising tool to help spread awareness. and finally I would like to bring a few other developers or developer onto this project to eliminate the risk that is being a solo developer. life happens but this project I feel is to important to risk a small accident or health issue stalling or stopping completion.

Who is on your team? What's your track record on similar projects?

its just me on this so far, however I do have a working demo at aicraftcyberforge.com I feel the best way to show if the work can be completed is to show you this is more than just a simple idea.

What are the most likely causes and outcomes if this project fails?

currently I have identified only 3 potential "fail" conditions, the first being operational baseline not being able to be sustained force me to take down the project due to running cost. the second being outreach/advertising not being able to be done diligently to spread awareness for the project to be able to be impactful. and the final fail condition I can see is a catastrophic injury or sickness preventing me from continuing the work. this last one I hope to be able to resolve almost instantly if I am able to secure some funding for the project.

How much money have you raised in the last 12 months, and from where?

so far the all the funding I have used has been self funded in total I have invested 1,600 in equipment 200 in hosting and domain. so a total of 1,800 invested into the project. this excludes all coding hours cost.