Loading video…
All breakouts
50.4Talking HeadHow-To hookAI Tools
Hook
Hi everyone, it's Sean, so today I wanna show you how to design and launch an AI deep research agent like a pro.
@sean_ai_stories
8.6K followers · usually 868 views a post
Views
43.0K
Likes
3.2K
Comments
39
Shares
355
Product placement
Kimi K2 (acts - used as an AI model for research, its performance is compared to other models, and its API is used for the research workflow), OpenAI (acts - used as an AI model for research, its API is used for the research workflow), Claude (acts - used as an AI model for research, its API is used for the research workflow), GPT-5 (acts - used as an AI model for research, its API is used for the research workflow), GPT-4o (acts - used as an AI model for research, its API is used for the research workflow), Google Docs (acts - the final report is exported to it), Google Drive (acts - the final report is exported to it), Vercel (acts - the frontend is deployed to it), Google Cloud (acts - the backend is deployed to it), GitHub (acts - the code is hosted on it), LangChain (acts - the deep research service is from this open source project), Moonshot AI (appears - its promotional activities are shown, its platform is mentioned as a place to use Kimi K2), Twitter (appears - a Twitter post from Kimi K2 is shown)
On-screen text
Design and Launch a Deep Research AgentnGoals:nRun Deep Research on a user querynGenerate a reportnExport to a Google DocnAI Agent Flow:nClarification -> Brief -> Supervisor -> Research -> ReportnTools:nWeb Search, Strategic think, MCPnResearch Conduct, Compress, CompletenAI Models: OpenAI, Claude, Kimi K2nDeep Research Agent FlownAI System Design Walk-throughnDeep Research Supervisor / RouternSupervisor Task Delegation to Researcher AgentsnAgent State ManagementnState Management for Agent, Supervisor, ResearchernClone, Configure, DeploynCreate Backend ReponDeep Research Backend SetupnDeep Research Frontend SetupnSetup Google Doc/Drive APInModel Comparison: Kimi K2, Claude 4, GPT-5, GPT-4onEval 1: Time Spent at Each StagenEval 2: Assessing Quality of ReportnFinal Comments
Original caption
You Can Learn Deep Research AI Agent Design & Launch In 25 Min | LangChain, OpenSource I'm Open Sourcing a Deep Research AI Agent System Design and Full Stack Codebase I built. This video shows you how I test Kimi K2 0905, Claude-4, GPT-5 on this end-to-end system of Deep Research, and export the research report to Google Docs. This isn't a mockup. It's a real functional product with a deep dive into its workflow. My public repos are built on top of the LangChain Open Deep Research functionalities, and introduced a streaming process to show the NextJS frontend how the FastAPI Python backend streams Deep Research steps hosted on Google Cloud (GCP). You can build anything. You can learn anything. 💪
More from @sean_ai_stories
More breakout videos from this creator.
