Agentic — June 07, 2026

Improve Your Agentic AI Trading With a Great Data Pipeline

The video explains how to build a multi-source data pipeline to improve agentic AI trading on platforms like Polymarket, using OpenAI's Codex as the agent. The creator walks through five data sources he has set up: Kalshi (competitor market data via API), Reddit sentiment analysis, Polymarket whale tracking (large bettors on-chain), X/Twitter searches, and Google/Chrome searches—all automated via a browser automation tool called Surf Agent. All collected data is compiled into a single master unstructured text file, which the AI agent then uses to identify and evaluate potential trades. As a live example, the pipeline identified a Formula 1 bet on Kimi Antonelli at 0.56 odds, and the position was already up 28% shortly after placement, demonstrating the approach's potential value.

How to Build Your First AI Agent in 10 Minutes (No Code)

This video is a sponsored tutorial showing beginners how to build an AI agent using Nexos.ai, a no-code platform, in about 10 minutes. The presenter walks through creating a morning briefing agent that automatically checks Gmail and Google Calendar each day at 7 AM and summarizes emails and schedule changes requiring attention. The build process involves typing a plain-language description of the desired agent, connecting Google integrations via OAuth, and selecting a delivery destination for results — no coding or server setup required. The video also highlights that Nexos.ai provides access to multiple AI models (ChatGPT, Claude, Gemini) under one subscription, and offers a discounted affiliate link with a 14-day money-back guarantee.

How to design a multi-agent system that skips the LLM

This video explores the architectural decision of when to use LLMs versus deterministic code in a multi-agent system, demonstrated through a real-world project of 1,000 simultaneous AI agents simulating a Las Vegas marathon. Casey Vest explains that marathon route planning uses classic computer science algorithms (Dijkstra/spine-and-sprout with Haversine-weighted edges) rather than Gemini, because the problem is NP-hard, requires exact precision (26.2188 miles), and is far more reliably and efficiently solved with deterministic, unit-testable code. However, AI (Gemini via AI Studio with code execution and Google Search grounding) was used during development to research and prototype those algorithms—just not at runtime. The core design principle is to apply LLMs where they add genuine value (ambiguous reasoning, research, language tasks) and skip them in favor of deterministic logic where precision, reliability, and cost efficiency matter.

Agentic AI Explained: Research AI Agents Are Powering the Next AI Revolution

The video explains what AI agents are and how they differ from standard AI chatbots, describing them as more like an "intelligent intern" that autonomously perceives information, reasons and plans, acts to generate outputs, and then learns and iterates — a cycle that can run for up to an hour on complex tasks. The presenter highlights that major platforms (ChatGPT, Gemini, Claude, Perplexity) already offer agentic modes, and that field-specific research agents exist across chemistry, biology, materials science, and machine learning. Using Claude's Co-Work agent as a hands-on example, the video demonstrates how it can conduct multi-search literature reviews, generate detailed reports with tables and references, create presentation slides from a PDF, and identify research gaps — all with optional human check-ins at key stages. The key takeaway is that agentic AI offers significantly richer, more autonomous outputs than simple chat interactions, making it especially powerful for academic and research workflows.

Best AI Video Agents in 2026 (Most Realistic)

The video compares three AI video generation agents—InVideo, HeyGen, and Pictory's Higgs Field Supercomputer—across three categories: credit transparency, UI clarity, and workflow quality, using an identical 30-second test prompt for each. Higgs Field wins the credit check category (9/10) for displaying costs before each step and requiring user approval, while InVideo (6/10) and HeyGen (5/10) both lack credit visibility, with HeyGen also wasting credits on ineffective regenerations. In UI clarity, Higgs Field scores 10/10 thanks to a fully functional stop button, whereas InVideo and HeyGen both receive 8/10 despite having real-time task visibility. For workflow quality, Higgs Field again dominates by generating multiple character/location options for user approval and using each previous video clip as a reference for the next, resulting in seamless, cinematic output—compared to InVideo's disconnected clips (6/10) and HeyGen's failure to apply edits or generate multiple avatars (4/10).