You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
LLM-ready JVM profiler — turns a JFR recording into a compact, source-attributed hot-path & allocation summary that AI coding agents can act on. In-process agent, MCP server, and JMH plugin; library + CLI.
Unofficial NVIDIA research dashboard for source-linked leadership coverage, NVDA vs. SPY market context, hypothetical CUDA roadmap scenarios, and affectionate Jensen Huang jacket satire. Built with Python and Streamlit; includes an interactive browser demo and safe local-only CLI.
RAG with citations: a Python FastAPI service where every answer cites the source document it came from, confidence is computed from citation coverage instead of model self-report, and cited ids missing from retrieval are flagged as hallucinations. FAISS retrieval, swappable LLM provider.
Offline Python tool for mapping claims to human-reviewed sources, preserving contrary evidence and review gaps. JSON, CSV and HTML reports. By Mohammed Teto — AI Reputation & Authority.
Replication package for “Illocutionary Explanation Planning for Source-Faithful Explanations in Retrieval-Augmented Language Models” (xAI 2026): code, data, prompts, and evaluation for chain-of-illocution prompting in textbook-grounded RAG.
sourcely is not a normal LLM app. It’s a RAG-based app that answers only from documents never from its own guesses. Every answer comes with the source, chunk index and the page number where it was found and one click opens the document at that page.
Self-hosted RAG demo with semantic search, cross-encoder reranking, and source attribution — built to show how the pieces fit together and what it takes to secure them.
Provider-neutral TypeScript library for evidence-grounded AI responses, with claim–evidence ledgers, streaming state, source integrity validation, Unicode-safe offsets, and React/Next.js integration.
Black-box investigation of Google AI Mode's misidentification of a real independent researcher, source attribution errors, and misinformation recurrence after correction. October 21, 2026.
CLI RAG system (with ChromaDB vector database) that consists of a relevance-gate guardrail against hallucination, grounded source-cited answers, and acceptance criteria validated through repeated, reproducible test runs for each of 5 success criteria rather than a single pass.