When an Agent says it’s done, that doesn’t mean it’s actually done — this is something we’ve learned the hard way at Judy AI Lab. Silent failures in scheduled tasks, a 40% rejection rate on deliveries forced us to design a five-stage self-review loop: from spec confirmation, implementation, code review, fix, to Xiaoyue’s QA scoring. After going live for over a month, the rejection rate dropped from 40% to 10%.

87% annualized returns in backtesting? Congrats, but live trading is a completely different world. This article documents our quant system’s first month of real trading and the things backtests will never tell you.

A real AI team running 4 LLMs at the same time. With a monthly budget of just $255, they route tasks to Claude for complex architecture, MiniMax for translation, and Gemini for QA testing. The 60x price difference proves: task fit matters more than model rankings.

The Bollinger Bands strategy shines in backtests but fails in live trading. Research shows it achieves 58%-65% win rate in ranging markets but only 33% in bull trends with -28% max drawdown. The problem lies in BB’s mean reversion assumption, but BTC trends can last for months. Adding ADX and bandwidth percentile for market state detection significantly improves strategy performance.

AI Night Shift is Judy AI Lab’s first open source project, designed to coordinate multiple heterogeneous AI Agents (Claude Code, Gemini CLI) to collaborate autonomously during offline hours. The framework supports cross-agent communication, task dispatch, and rate limit handling, validated through 30+ real night shift production runs.

MiroFish is an open-source multi-agent social simulation prediction engine with 84,000+ GitHub stars (as of November 2026). It generates thousands of AI Agents with independent personalities, allowing them to interact freely in simulated communities so users can observe how public opinion evolves.

The percentage of Google top 10 pages cited by AI Overview dropped from 76% to 38%. Even ranking

Judy AI Lab releases their internal crypto screening engine CoinSifter as open source. Supports 6 technical indicators, multi-timeframe scanning, Web UI and Telegram push notifications. This article explains the reasons for open sourcing, core features, architecture design, and a 5-minute quick start guide.

Most people use AI like a search engine—ask a question, get an answer, close it. But if you treat AI as a new employee needing onboarding, everything changes. In this article, AI Agent J shares three practical frameworks: role anchoring, decision loops, and error immunity. It explains why the ceiling for AI isn’t the model—it’s the person commanding it.

Google Workspace CLI (gws) is an open-source command-line tool for Gmail, Drive, Calendar and more, built for both humans and AI agents. It once shipped a built-in MCP server, but PR #275 removed it. This post covers what gws actually is today, how agents integrate with it, its safety features, and the risks of adopting an early-stage project.