IGIgor Ganapolskyinai-agent-blog-12345.hashnode.dev·3d ago · 2 min readWhy Agent Approval Must Happen Before the Tool Call ExecutesRemote access and human control are different engineering problems. A mobile client can start an agent, stream tokens, and send another prompt. That is useful remote access. It does not create an enfo00
IGIgor Ganapolskyinai-agent-blog-12345.hashnode.dev·Jul 5 · 3 min readMemory Is Not Enforcement: Why AI Agents Need Pre-Action ChecksMost agent teams already have memory somewhere. A repo note. A chat history. A vector store. A wiki. A status bar. A dashboard full of things the agent should remember next time. That helps the ne00
IGIgor Ganapolskyinai-agent-blog-12345.hashnode.dev·Jun 17 · 2 min readAI agents need source-of-truth gates, not better vibesI shipped `thumbgate@1.27.8` today because a recurring agent failure kept showing up in the real world: Agents do not only make bad tool calls. They also make confident business claims after weak or p00
IGIgor Ganapolskyinai-agent-blog-12345.hashnode.dev·Jun 12 · 3 min readYour AI Agent's Mistakes Are a Free Preference Dataset. Stop Deleting Them.You correct your AI coding agent dozens of times a day. "No, not that file." "Don't force-push to main." "That function doesn't exist — you hallucinated it." Each of those corrections is a perfectly l00
IGIgor Ganapolskyinai-agent-blog-12345.hashnode.dev·Jun 10 · 3 min readYour AI agent forgets your repo every session. Give the repo a brain.84% of developers now use or plan to use AI coding tools. Only about a third fully trust the output — and that trust has been falling since 2023. The usual explanations are hallucination and skill. I 00
