KILLBEROS
Risk-First (Educational · Fully simulated)
Measure the danger before the gain.
Most small investors don't lose for lack of ideas they lose for lack of risk discipline: they over-trade, double a losing position, and over-size out of fear or greed. KILLBEROS gives anyone the risk discipline of an investment fund, on a modest $1,000 budget ($500 in play, €500 kept untouchable). 100% educational, fully simulated no real money, no broker, no orders ever placed.
• Risk Planner — real position sizing on any instrument: size, risk, reward-to-risk, computed live.
• The 7 Gates — every trade must pass 7 risk checks; break one (e.g. a reward:risk below 2:1) and it's BLOCKED. The danger is measured first.
• Market Regime — live market "weather" (real Fear & Greed + real crypto prices) decides which strategies are even allowed to play.
• AI Advisor + a mistake "pokédex" — so the same error never repeats.
• Journal — log real trades, see your true edge: win rate, expectancy, drawdown (real backend, not a mockup).
Innovative Workflow)
KILLBEROS was designed and built in Figma Make, then reviewed live and iterated, version after version (V2 -> V43), by an AI agent "Cowork Brain" working through Figma's MCP.
But the real innovation is HOWdecisions were made: every significant choice (doctrine, a risk gate, a strategy, the UI) was debated by a standing panel of 11 AI "voices" that the agent embodies in one head — 5 trader profiles (macro, chartist, a risk/Aladdin voice with veto power, a short-term "gambler", and a long-term allocator) + 5 engineer profiles (backend, UX, API, security, data).
Risk speaks first, the angles are confronted, a decision is made, and the dissenting view is written down. (Junior) signs off on anything irreversible. It's a small AI team with a real method not a single prompt. Build-in-public: every version and decision is journaled.
And After :
KILLBEROS is the educational tip of the iceberg. Behind it, a far more powerful engine the full GambleRisk is in the works: the same risk-first doctrine, but it paper-trades itself for weeks to PROVE its edge, market regime by market regime, before anything else. Still risk-first, still discipline-before-gain. This is just the beginning.
Team : Junior Odika (+ Cowork Brain, AI collaborator).
URL :
https://www.figma.com/make/qnXCf663p4TkScVfXV0KiO/KILLBEROS-Web-App-Design?p=f&t=UGdrmbyAWU8Pto88-0
https://hasty-gothic-71393447.figma.site/
4
4
297
I design AI automations that take repetitive work off a team's plate. This one is a customer-care pipeline: it captures every email and web-form request, uses OpenAI to classify the message, gauge its urgency and read the customer's tone, then routes it automatically escalating urgent cases to a manager, sending instant acknowledgments, alerting the team on Slack and logging every ticket to Google Sheets. Built with n8n: reliable, secure, and live around the clock.
1
51
I build AI-powered automations that run a business while you sleep. This project is an autonomous client agent: it reads incoming emails and website forms in real time, uses Google Gemini to classify each message, score its priority and read the customer mood, then drafts a ready-to-send reply. Critical cases are escalated instantly to email and Slack, standard ones get an automatic personalized answer, and every interaction is logged to Google Sheets. Built with n8n fast, secure, deployed.
1
23
Strike Vault is a combat-sports e-commerce site I designed and built from scratch. Cinematic front-end (neon title effects, scroll animations), a multi-category product catalogue, and a full cart with secure Stripe payment — all in Next.js + TypeScript. I also built business automations with n8n: contact/support, customer database, newsletter and product analytics. From design to deployment, a fast, secure store ready to sell. Built for fighters, engineered to convert.
0
16
GambleRisk is an algorithmic trading bot I built from scratch, running live on Binance & Capital.com (http://Capital.com). It watches markets in real time, tracks macro news, and merges multiple data sources into one consensus score. Every signal passes a risk engine: stop-loss, take-profit, position size and confidence are auto-calculated. It even dialogues with Claude AI before deciding — and must survive a 28-day simulated incubation before earning real execution. Intelligence, discipline, zero blind trust.