EIEshan Inamdarineshaninamdar.hashnode.dev00Agentic AI Deep Dive — Failure Modes: What Goes Wrong and Why3d ago · 11 min read · Where We Left Off Part 7 was about deliberate attacks — prompt injection, trust boundary violations, agents with too much permission. The threat model was an adversary trying to manipulate your agent.Join discussion
EIEshan Inamdarineshaninamdar.hashnode.dev00Agentic AI Deep Dive — Part 7 Security & TrustMay 23 · 12 min read · The Story So Far — Full Series Recap Six parts in. Here is what we've built up. Part 1 — LLMs have four fundamental limitations: knowledge cutoff, no private data, statelessness, no actions. RAG fixedJoin discussion
EIEshan Inamdarineshaninamdar.hashnode.dev10Agentic AI Deep Dive — Multi-Agent SystemsMay 20 · 10 min read · Where We Left Off In Part 5 we solved the memory problem. A well-designed agent uses four memory types — in-context for the active task, external for large persistent stores, episodic to learn from paJoin discussion
EIEshan Inamdarineshaninamdar.hashnode.dev00Agentic AI Deep Dive — Memory: How Agents RememberMay 19 · 11 min read · Where We Left Off In Part 4 we gave agents hands. Tools are how an agent reaches outside itself — into APIs, databases, browsers, file systems. The LLM describes an action, the harness executes it, thJoin discussion
EIEshan Inamdarineshaninamdar.hashnode.dev00Agentic AI Deep Dive - Tools: How Agents Act on the WorldMay 18 · 12 min read · Recap: The Story So Far Three parts in. Here's where we stand. In Part 1 we established why LLMs alone break for real work. Four fundamental limitations — knowledge cutoff, no private data, statelessnJoin discussion
