Built and secured a full AI system in four weeks with no coding background

Achievements

- Numerous, detailed in attached showcase document

- Everything described in showcase document — designing, building and testing the AI System — was carried out over a period of roughly four weeks, from the start of the project to today.

Primary Tools

- Claude Chat (Opus 4.8, 5 and Fable 5)

- Claude Code (Opus 4.8, 5 and Fable 5)

- Terminal sessions (on Macbook Pro)

- Pinecone

- Python code (prompted by me, written by Claude)

- Swift code (prompted by me, written by Claude)

Methodology

- The security wasn't assumed — For testing, it was deliberately attacked and then proven. A full abuse battery was run and passed 100%; in every case the system behaved correctly and leaked nothing. The tests included multiple testing against various prompt injection, such as (a) Instruction override, (b) System prompt extraction, (c) Role-play jailbreak and (d) Data exfiltration (and others described in the document.

- Multi layer rate limiting to protect against abuse.

- The project runs on a strict “never assume — verify by state” method: nothing is declared done until checked against reality, every change is version-controlled and logged, and multiple independent gates must pass before anything reaches users.

- Every decision reasoned and recorded, every change version-controlled, and every claim verified against the real system rather than assumed.

- Claude code session management with running artifacts session to session as learnt from the course. (The Claude project so far has 73 chat sessions).

Learnings

- When this project began, my experience of AI was limited: some casual use of ChatGPT for simple chat, no hands-on work with Claude, and no exposure to vector databases such as Pinecone.

- The AI Masterclass course changed that, giving both a solid grounding in the fundamentals of AI — its key components and features — and, just as importantly, a sense of what is genuinely possible. I complemented the course with my own research to deepen those foundations.

- I have a strong technical background, but by profession I am not a software developer, programmer or coder. All that is described in the attavhed showcase document was achieved by pairing that technical grounding, IT security knowledge, and a disciplined working method with AI assistance; making use of all the tools and techniques learn't on the course.

- Everything described in attached document — designing, building and testing the AI System — was carried out over a period of roughly four weeks, from the start of the project to today.

You can try out the new AI system via the organisations App, as a beta tester, available in the Android Play store and Apple App store.

Details of what app to get (free) and how to access the beta tester area are all provided in the "TrueSource AI - Showcase" document attached.

If you have any questions please email me. Same email that is registered with Mindvalley.

I have submitted a detailed document as an attachment that includes the following sections

1. Background and context

2. What the project is

3. Summary of artifacts and outputs

4. Features for users

5. Features for the administrator

6. Security - designed layer by layer

7. Beta Tester's area

8. HOW YOU CAN TRY OUT THIS AI SYSTEM AS A BETA TESTER

9. Other things worth showcasing

10. How it was built

11. Current status