Like this post? Sign up or log in to like, comment, and connect with this founder.
Post Details
Abheejit Deka
@devilgod
I am 15, and I love creating complex systems, solving riddles, and writing digital stories. Whenever I tried to use standard AI tools to help me map out massive narrative arcs or technical structures, I ran into a massive wall: context loss. Standard chat apps are just simple wrappers—they completely forget your variables, character states, and rules the second you close the tab or hit a certain length.
I decided to stop looking for a tool and just build one myself.
Phase 1: Designing the Architecture
I didn’t want to build a generic chatbot. I wanted a developer workspace. I spent time sketching out layout concepts and building an intentional, multi-layered architecture:
Layer 2 Matrix (The State Layer): I coded a persistent environment layer that tracks active project variables, character sheets, and structural rules across sessions.
The Review Pipeline: I integrated a backend pipeline that runs automated critique loops, recursively checking text or code blocks against specific rules.
Phase 2: Solving the Token/API Bottleneck
While building and tweaking the UI with Python, I quickly ran into a classic engineering hurdle: running out of live API tokens mid-testing.
Instead of letting rate limits stall my progress, I engineered a Local Mock-Testing Safe Mode. I wrote an asynchronous mock engine that completely isolates live networks during interface development. It simulates network latency, intentionally injects rule violations for the review pipeline to catch, logs test data straight into a local SQLite database, and populates the telemetry dashboard. This lets me test UI changes, CSS tweaks, and database routing endlessly without burning a single token or paying a dime.
Phase 3: Bringing it Together
Right now, the full backend, modular database workflows, and real-time diagnostic telemetry panel are up and running smoothly. Everything is fully functional and modularly built.
I have handled 100% of the engineering and backend execution myself—but I know my strengths. Marketing and community outreach aren't my things. I am sharing this story on Teen Founders India to find a dedicated growth and marketing co-founder around my age to help take this engine to the public!
I decided to stop looking for a tool and just build one myself.
Phase 1: Designing the Architecture
I didn’t want to build a generic chatbot. I wanted a developer workspace. I spent time sketching out layout concepts and building an intentional, multi-layered architecture:
Layer 2 Matrix (The State Layer): I coded a persistent environment layer that tracks active project variables, character sheets, and structural rules across sessions.
The Review Pipeline: I integrated a backend pipeline that runs automated critique loops, recursively checking text or code blocks against specific rules.
Phase 2: Solving the Token/API Bottleneck
While building and tweaking the UI with Python, I quickly ran into a classic engineering hurdle: running out of live API tokens mid-testing.
Instead of letting rate limits stall my progress, I engineered a Local Mock-Testing Safe Mode. I wrote an asynchronous mock engine that completely isolates live networks during interface development. It simulates network latency, intentionally injects rule violations for the review pipeline to catch, logs test data straight into a local SQLite database, and populates the telemetry dashboard. This lets me test UI changes, CSS tweaks, and database routing endlessly without burning a single token or paying a dime.
Phase 3: Bringing it Together
Right now, the full backend, modular database workflows, and real-time diagnostic telemetry panel are up and running smoothly. Everything is fully functional and modularly built.
I have handled 100% of the engineering and backend execution myself—but I know my strengths. Marketing and community outreach aren't my things. I am sharing this story on Teen Founders India to find a dedicated growth and marketing co-founder around my age to help take this engine to the public!
11:55 AM · Jun 29, 2026
0
viewsLog in to join the conversation
Log in