@dantekakhadze

A2A + MCP: two protocols, one interoperability layer

1h ago · 6 min read · The agentic AI world has been quietly converging on two protocols. MCP (Model Context Protocol), originally from Anthropic, now under the Linux Foundation's Agentic AI Foundation, handles how agents connect to tools and data. A2A (Agent-to-Agent), la...

MCP security: every unvetted server is an attack surface you chose to ignore

1h ago · 7 min read · MCP adoption is moving fast. The official Python and TypeScript SDKs now see over 97 million monthly downloads. Anthropic donated the protocol to the Linux Foundation's Agentic AI Foundation in Decemb

Full autonomy is a trap. Human-in-the-loop is the production architecture.

3h ago · 6 min read · The pitch is always the same: "Our AI agent handles everything end-to-end, no human intervention required." Sounds great in a demo. In production, full autonomy is how you get agents that process refu

Agent memory shouldn't be a hack. Here's what a real implementation looks like.

3h ago · 6 min read · Every agent framework has a memory story. Most of them amount to "we append previous messages to the context window." Some get fancier with vector stores for long-term recall. A few use summary chains

Why 40% of AI projects fail (and it's not the model's fault)

3h ago · 5 min read · You've seen the stat. Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027 due to escalating costs, unclear business value, or inadequate risk controls. Leadership blames t