Local AI models. An interface that makes sense.
Explanations and highlights. You stay in control.
macOS first. iPhone and iPad support planned.
Designed to understand visible context through OCR.
On-device AI. iCloud sync for chats and files is planned.
Couldn’t load the reveal. Watch the video
MACHINE LEARNING · MARKET RESEARCH
An open-source engine that turns public market data into tested models and paper signals.
Public Coinbase and Kraken candles. Checked together before they become a research dataset.
INSIDE THE ENGINE
Ares checks the input before it trusts the output. Change the feed below. See what gets through.
Fresh, complete data passes every check.
PASS / VALIDATED INPUTIllustrative scenarios, including simulated recovery. No live exchange connection or orders.
EVIDENCE OVER INSTINCT
Learn from the past. Leave a gap. Test on what the model hasn’t seen.
Ares uses chronological evaluation with fees, slippage, and doubled-cost stress. Models earn their place through evidence.
Read the research protocolWindow 01. Train on the earliest data, then evaluate on the next unseen period.
Select a window to follow the expanding training history.BUILT TO BE EXAMINED
Explore the research. Inspect the decisions. Make it better.
Gabriel Williams / Founder & software engineer
I turn ideas into software with purpose, character, and clear limits.
Arjia is my independent practice in applied AI and product engineering. I build websites, custom software, and user interfaces, connect the systems behind them, and automate workflows with AI.
Understand the problem, constraints, and consequence.
Build the interface and the system behind it.
Publish the evidence, decisions, and limits.