Post by Rakesh Gohel

While everyone is obsessed with MCP and A2A in AI Agents These Open-source projects are quietly reshaping AI 👇 1/ Deepseek OCR: https://lnkd.in/gFwPAZqB Vision+LLM fusion, insane 5–8x token savings, just $0.03 per 1M tokens! Context compression you can actually afford. 2/ Microsoft Agent Lightning: https://lnkd.in/giz-zJ9d Unlock multi-agent superpowers by injecting RL into any system, no rewrites needed. 3/ Anthropic Agent Skills: https://lnkd.in/ggUMshc2 Best replacement for prompt engineering, allows agents to follow the context better with virtual skills testing and implementation. 4/ Toon (Token-Based Object Negotiation): https://lnkd.in/gZkDzwqD Minimal-token, human-readable data exchange for LLMs. 5/ Google AP2 Protocol: https://lnkd.in/gnZexJ8r Next-gen payments: lets agents securely buy anything, cards, crypto, bank; no human click, full audit trail. 6/ Agentic Commerce Protocol: https://lnkd.in/g_9_Wqi7 Open standard to let agents shop for you anywhere. Direct, secure, instant checkout with any processor 7/ Microsoft Markitdown: https://lnkd.in/g65-EaMv Converts docs to Markdown for LLM pipelines; making agent input simple, rich, and context-packed. 8/ IBM Docling: https://lnkd.in/grf3y2sv Hallucination-free doc parser. Handles PDF, Office, images; clean input for GenAI and RAG. 9/ Google Coral NPU: https://lnkd.in/gAnB8hXF Tiny but mighty open-source RISC-V AI chip for wearables; efficient, ultra-low power, ML inferencing. 10/ Sapientic Hierarchical Reasoning Model: https://lnkd.in/gFqtFpES Small but genius: 27M params, beats giants on AGI reasoning. What other innovations do you think are shaping the future of AI? Let them know in comments so that everyone can learn 👇 Save 💾 ➞ React 👍 ➞ Share ♻️ & follow for everything related to AI Agents