https://alkhaldieid.github.io/blog/ Most teams can access the same models. My work is deciding what can be trusted, what must be verified, and how AI becomes an auditable system inside regulated Arabic-English workflows. en https://alkhaldieid.github.io/blog/posts/rpa-brokers-control-before-automation/ https://alkhaldieid.github.io/blog/posts/rpa-brokers-control-before-automation/ Mon, 18 May 2026 07:44:49 GMT Broker automation needs a control layer, not another set of screens wired together. https://alkhaldieid.github.io/blog/posts/fine-tuning-flops-ai-governance/ https://alkhaldieid.github.io/blog/posts/fine-tuning-flops-ai-governance/ Wed, 13 May 2026 10:04:34 GMT Compute accounting is becoming a governance record, not only an infrastructure metric. https://alkhaldieid.github.io/blog/posts/mcp-context-management/ https://alkhaldieid.github.io/blog/posts/mcp-context-management/ Wed, 13 May 2026 09:53:10 GMT MCP should be understood as an interface discipline: it makes context and tools explicit enough to govern. https://alkhaldieid.github.io/blog/posts/mcp-production-memory-context/ https://alkhaldieid.github.io/blog/posts/mcp-production-memory-context/ Wed, 13 May 2026 09:49:53 GMT MCP is useful only when teams treat context, tools, and session state as governed interfaces. https://alkhaldieid.github.io/blog/posts/mcp-agents-session-integrity/ https://alkhaldieid.github.io/blog/posts/mcp-agents-session-integrity/ Wed, 13 May 2026 08:53:30 GMT The hard problem is not calling one tool; it is preserving meaning across a sequence of stateful tool interactions. https://alkhaldieid.github.io/blog/posts/mcp-agent-infrastructure/ https://alkhaldieid.github.io/blog/posts/mcp-agent-infrastructure/ Wed, 13 May 2026 08:49:04 GMT MCP turns ad hoc agent integrations into contracts that can be tested, governed, and changed. https://alkhaldieid.github.io/blog/posts/gulf-enterprise-agents-closed-loop-control/ https://alkhaldieid.github.io/blog/posts/gulf-enterprise-agents-closed-loop-control/ Sun, 03 May 2026 09:57:25 GMT For Gulf enterprise agents, reliability depends on validating actions before they touch sensitive systems. https://alkhaldieid.github.io/blog/posts/metered-reasoning-for-ai-agents/ https://alkhaldieid.github.io/blog/posts/metered-reasoning-for-ai-agents/ Fri, 01 May 2026 08:18:57 GMT Agent quality will depend less on constant reasoning and more on deciding when extra reasoning is worth the cost. https://alkhaldieid.github.io/blog/posts/insurance-ai-reasoning-grounded-retrieval/ https://alkhaldieid.github.io/blog/posts/insurance-ai-reasoning-grounded-retrieval/ Mon, 27 Apr 2026 10:34:59 GMT In insurance, high-risk errors usually come from weak grounding rather than weak language generation. https://alkhaldieid.github.io/blog/posts/gulf-arabic-rag-context-assembly/ https://alkhaldieid.github.io/blog/posts/gulf-arabic-rag-context-assembly/ Mon, 27 Apr 2026 05:33:47 GMT Arabic RAG failures frequently come from weak context assembly rather than weak embeddings. https://alkhaldieid.github.io/blog/posts/humility-verification-llm-behavior/ https://alkhaldieid.github.io/blog/posts/humility-verification-llm-behavior/ Mon, 27 Apr 2026 05:14:38 GMT What looks like model humility is often an architecture that forces the system to pay for evidence. https://alkhaldieid.github.io/blog/posts/llm-confidence-evidence-loop/ https://alkhaldieid.github.io/blog/posts/llm-confidence-evidence-loop/ Sun, 26 Apr 2026 09:43:39 GMT LLM confidence is less about model size and more about whether the system is forced to verify. https://alkhaldieid.github.io/blog/posts/why-rag-fails-when-retrieval-is-not-evaluated/ https://alkhaldieid.github.io/blog/posts/why-rag-fails-when-retrieval-is-not-evaluated/ Sat, 25 Apr 2026 15:13:40 GMT Many RAG failures are diagnosed as model failures even when the real error happened before generation. https://alkhaldieid.github.io/blog/posts/saudi-enterprise-rag-budget/ https://alkhaldieid.github.io/blog/posts/saudi-enterprise-rag-budget/ Sat, 25 Apr 2026 15:10:49 GMT The expensive part of RAG is not always the model. It is the absence of retrieval discipline. https://alkhaldieid.github.io/blog/posts/why-ai-teams-still-need-python/ https://alkhaldieid.github.io/blog/posts/why-ai-teams-still-need-python/ Sat, 25 Apr 2026 14:54:09 GMT No-code orchestration can be useful at the boundary of a system, but it should not become the center of an AI engineering practice. https://alkhaldieid.github.io/blog/posts/llms-legal-compliance-institutional-memory/ https://alkhaldieid.github.io/blog/posts/llms-legal-compliance-institutional-memory/ Fri, 24 Apr 2026 15:30:57 GMT Compliance systems need institutional memory: how similar cases were interpreted, escalated, and resolved. https://alkhaldieid.github.io/blog/posts/arabic-llm-context-evaluation/ https://alkhaldieid.github.io/blog/posts/arabic-llm-context-evaluation/ Fri, 24 Apr 2026 15:15:14 GMT Arabic model quality should be judged by contextual behavior, not by fluency alone. https://alkhaldieid.github.io/blog/posts/arabic-llm-semantic-evaluation/ https://alkhaldieid.github.io/blog/posts/arabic-llm-semantic-evaluation/ Fri, 24 Apr 2026 14:32:27 GMT Arabic evaluation cannot stop at surface overlap; it must test whether meaning survives dialect, formality, and domain language. https://alkhaldieid.github.io/blog/posts/self-evolving-agents-reliability/ https://alkhaldieid.github.io/blog/posts/self-evolving-agents-reliability/ Thu, 23 Apr 2026 19:04:31 GMT Adaptive agents require stronger observability because the system that runs today may not behave exactly like the one tested yesterday.