Articles

Multi-Agent AI Platform

An insight into our experience build a multi-agent AI platform for research and documentation. The article will provide insights into the mistakes made and the lessons learnt along the way

This paper is an honest account of building a real agentic AI platform from scratch. It is not a framework pitch or architecture guide. It is a record of what worked, what failed, what should have been built first, and what would be done differently. The system built is Multi-Agent Studio, a Streamlit web platform hosting AI agent modules.

Module 1 is a Research Assistant: a 9-agent pipeline that takes a user topic and produces a polished research paper end to end. The pipeline runs: Planner, Researcher, Critic, Writers A and B in parallel, Debate Judge, Fact Checker, Judge, and Editor. Built over 11 sessions with Claude Code. Stack: Python, Streamlit, LangGraph, Gemini API, Groq API, Tavily API, and Exa API. Every lesson below came from real failures in this build.

The goal was a working multi-agent research pipeline, not a toy demo. Module 1 had to accept a topic, search the web, critique the sources, write a paper with two competing writers, judge the debate, fact-check the winner, and edit the final output. Nine agents, real APIs, live web search, deployed on Streamlit Community Cloud. The purpose was to learn agentic design by building something with real failure modes.

multi-agent-studio

multi-agent-studio.pdf

212 KB

This illustration depicts some of the possibilities over the next 10 years.

AI - the Next Decade

This is my prediction, not a claim of certainty: where AI is headed over the next ten years, based on what I see happening today and what builders and users of the technology are already saying. What's already decided, and what genuinely isn't.

This illustration shows how Ai and human judgement need to go hand in hand

AI, Accountability, and the Future of Human Judgment

AI accountability, human judgment, and the capability gap most companies never measure. This insight paper covers AI disclosure, agentic AI oversight, skill erosion, and the productivity numbers that do not add up. A clear standard for honest, defensible AI use and governance.

Illustration showing how AI skills are transferable to professional life

Ten Ideas to Build Personal AI Habits Before Work Adoption

Stop treating AI like a search engine. From restoring photos to planning dinner, these ten simple exercises build the "elite" skills you need for work. Master AI through habits you actually care about and then bring that power to your work use cases. Read more to bridge the gap!

The illustration reveals how to use multiple AI tools to enhance your content

The Multi-Model Blueprint: How to Use Multiple AIs to Create Superior Content

Stop hitting the invisible ceiling of single-AI workflows. This blueprint reveals how to manage a team of AI experts—mixing tools like Claude, ChatGPT, and Perplexity to shatter limits and create bulletproof, elite-tier content. Stop prompting; start building your production line.