AI Workflows & Agentic Architecture Directory

Kimi K3 self-hosted coding pipeline: deploy Moonshot's 2.8T open-weight MoE model locally with vLLM. Complete guide covering hardware requirements, Open Interpreter setup, Kimi Delta Attention, benchmarks vs GPT-5.6/Opus...

PrismML Bonsai 27B on-device AI guide. Deploy 27B model at 3.9 GB on iPhone 17 Pro or 5.9 GB on laptops. Apache 2.0, tool calling, vision, reasoning. 163 tok/s on RTX 5090.

ccshare multiplayer Claude Code: real-time collaborative sessions with shared thread state. Setup guide, conflict resolution, parallel subagents, team review.

AI copyright compliance audit guide — training data inventory, output log sampling, provenance tracking, and litigation hold automation for enterprises.

YAGNI (Product Hunt #10, July 16, 2026) introduces agent Teams managed like human employees: Responsibilities (plain language), a Number (single metric), Commitments (end dates). Trust progression: Training → Supervised...

OpenAI Codex PR #26210 (June 5, 2026) encrypts Multi-Agent V2 message payloads with AES-256-GCM, removing operator visibility into agent-to-agent instructions. Developers filed GitHub issues #28058 and #32753. Forced enc...

Claude Overlay (MIT, Product Hunt #24 July 16, 2026) is a floating screen-aware Claude Code desktop overlay for Windows that captures all monitors. Uses claude-agent-sdk, Pillow ImageGrab, Tkinter. No API key — uses Clau...

GPT-Red (OpenAI, July 15, 2026) is an LLM super-hacker trained via self-play RL that finds prompt injection vulnerabilities humans miss. 84% attack success vs 13% human on indirect prompt injection. Discovered Fake Chain...

Inkling (Thinking Machines Lab, 975B MoE, Apache 2.0) vs GLM 5.2 (Zhipu AI, top coding/reasoning) vs Nemotron 3 Ultra (NVIDIA, 550B): benchmark comparison across SWEBench, Terminal Bench, MCP Atlas, HLE, AIME 2026, and c...

OpenAI Codex Multi-Agent V2 (PR #26210, June 5, 2026) encrypts inter-agent message payloads with AES-256-GCM for GPT-5.6 Sol and Terra, breaking operator visibility into delegated agent instructions. Developers filed iss...

GPT-Red (OpenAI, July 15, 2026) is an automated red-teaming LLM trained via self-play reinforcement learning that discovers prompt injection vulnerabilities humans miss. It achieved 84% attack success vs 13% for human te...

Inkling is Thinking Machines Lab's 975B MoE open-weights model (Apache 2.0, July 15, 2026) with 41B active parameters, 1M context, native multimodal input, and controllable thinking effort (0.2-0.99). Founded by former O...