Articles | Green Olive Tech

Articles

Insights & Announcements

Tutorials, product updates, and engineering perspectives from the Green Olive Tech team.

April 4, 2026·16 min read·Mohammed Ahmed

30 Markdown Files Won't Save You: What Actually Works When Building AI Agent Teams

I built an entire product using 8 AI agents — 232 tests, 34 design assets, 9 content pieces. Here's what worked, what failed, and why governance beats file trees.

ai-agentsclaude-codemulti-agentgovernancedeveloper-toolsagent-teams

March 31, 2026·13 min read·Mohammed Ahmed

36% Fewer Wasted Cycles: Empirical Proof That Structured Governance Makes AI Agents Measurably Better

A controlled A/B benchmark across 5 production codebases proves that structured governance files reduce AI agent discovery overhead by 36%, cut wall-clock time by 17%, and eliminate the single largest source of wasted compute in agentic AI workflows.

AI AgentsClaude CodeAI GovernanceDeveloper ProductivityLLM Engineering

March 22, 2026·8 min read·Mohammed M. Ahmed

I Stopped Using Claude Code as a Tool and Built It Into an Execution Platform

A practitioner's account of building a governed, self-improving AI execution environment around Claude Code — 8 layers, 5 hooks, 12 skills, and a quality ceiling that rises on its own.

claude-codeai-engineeringdeveloper-toolsai-governancesoftware-architecture

March 3, 2026·13 min read·Green Olive Tech

The End of AI Guesswork: How CodePrizm Delivers Validated Codebase Intelligence

When AI analysis tools hallucinate file paths and fabricate vulnerabilities, the consequences are not academic. CodePrizm was built to solve this with a validation-first architecture.

code-analysisAIsecuritydata-platformsvalidationenterprise

March 3, 2026·12 min read·Mohammed M. Ahmed

How We Achieved 10x LLM Throughput by Migrating from Ollama to vLLM

We migrated a production text processing pipeline from Ollama to vLLM and saw throughput jump from ~67 ops/min to over 675 — a verified 10x improvement through continuous batching, batch prompting, and right-sized model selection.

vllmollamainferencellmperformancegpuproduction