Welcome to AI Academy!

AI benchmarks, AI Newsletter, Youtube videos and AI Agents

Subscribe to the AI Newsletter

Humanity's Last Exam Results

This benchmark measures AI models/LLMs accuracy, % of correct answers, on a human‑graded reasoning exam, higher is better.

How AI Agents Work

Agents are software helpers that break a goal into steps. Hover a panel and keep your mouse still to watch the flow. Move your mouse to pause; stay still to resume. Leave the panel to reset.

Basic Agent

One agent. Simple, linear flow.

  • 1 agent
  • Minimal pipeline
  • Deterministic steps
User prompt
Agent plans
LLM generates
Response

Intermediate Agent

Orchestrator coordinating three agents.

  • 3 agents + orchestrator
  • Structured outputs
  • Quality check
User request
Orchestrator routes
Research Agent
Writer Agent
Reviewer Agent
Final answer

Advanced Agent

Multiple agents with tools, memory, and guardrails.

  • 6+ agents
  • Guardrails & evaluation
  • Databases & tools
User goal
Orchestrator + Planner
Research Agents
Tool/Browser
Retrieve from DB
Guardrails/Evaluator
Synthesis Author
Finalized answer

YouTube Videos

Photo of Diego Cabezas
About the creator:
Diego Cabezas

AI Senior Associate and Data Scientist at JPMorganChase, based in Orlando, Florida. Originally from Quito, Ecuador, Diego has a deep passion for Artificial Intelligence, having transitioned from a background in digital marketing and business management into AI research and development. He specializes in building AI solutions that drive operational efficiency, including LLM-powered chatbots, speech systems, and predictive models. Diego also runs the AI Academy YouTube channel, where he teaches the latest AI tools and concepts, and was featured in Forbes Ecuador’s Under 30. His goal is to become a leading voice in AI research and innovation.