Hi, I'm Danila.

I started this project because I was tired of adapting my workflow to the T‑Invest terminal and splitting my attention across its screens. I wanted to decide how portfolio data, P/L, risk, and market context should be calculated and presented. What began as a personal dashboard has become a separate .NET and React production system with its own database, background jobs, and broker synchronization.

The system now produces technical, fundamental, and combined AI analysis. Instead of a single line, it stores several scenarios with probabilities, W1–W4 waves, entry zones, targets, and segment-specific stops. A forecast can be overlaid on the broker price chart and reopened later as an immutable archive against the realized candle path. The closest branch is identified automatically while its original probability remains visible.

AI still does not place trades for me. I use the system as a research and control layer: it tracks forecast quality, separates target-touch from the stricter target-before-stop result, and sends generated signal charts to Telegram. As verified observations accumulate, I get a clearer picture of when a model's conclusion is useful for entry or risk control—and when it should be ignored.

Sometimes I return to a book after a long break and realize that I have forgotten its characters, events, and terminology. Ordinary search is risky because it does not know how far I have read and can reveal a spoiler from a future chapter. I built a reader whose reference material is always limited to the current position.

The service processes chapters in order and maintains a Wiki of characters, places, events, items, and abilities. When a new chapter is analyzed, the model receives previously known entities and facts, which improves alias resolution and prevents duplicate cards for the same character. Two background workers can process different books in parallel while preserving chapter order within each book.

The idea can now be tried without an account using the public Harry Potter and the Methods of Rationality demo: reading and the position-aware Wiki are available while the source book and private user data remain isolated. I use DeepSeek as the most accessible AI environment for this task, but the architecture can support other models and compare their extraction quality.

I wanted a Quoridor match to start without finding the box, clearing a table, or asking the other player to install another app. So I built one online game that works both as a regular website and as a Telegram Mini App.

The game engine runs on the server and validates every move consistently: pawns cannot pass through walls, and a new barrier may never block every path to the goal. Players can join through an invitation or the shared matchmaking queue; interrupted sessions reconnect, while completed games remain available as animated replays.

On mobile, wall placement is a clear two-step interaction with preview and confirmation. Around the match itself, the product already includes profiles, statistics, leaderboards, an online counter, and a Telegram bot with group rankings. The project is live and continues to grow as a small platform for quick competitive games.

A contact list answers “how do I reach this person?” but says almost nothing about context. I want to remember where we met, which people, places, and events connect us, how a relationship changed, and which circle a particular story belongs to. That need became Social Connections.

Instead of a flat address book, the system builds an interactive graph. People, places, and events become nodes; directed and parallel relationships can carry roles, notes, dates, and media. Different parts of life remain in separate workspaces, while selected spaces can be combined into a read-only map without merging their source data.

A large graph still has to be understandable. The interface filters entities and relationship types, expands neighbors on demand, and shows a point in time. Hovering a relationship keeps exactly that edge and its two endpoints in focus while dimming everything else. This is a private beta I am developing as personal memory for social context—not as another social network.