Skip to content
Cranking out Good Code
Go back
Six Talks at AIMUG: Building Agents That Work
Talk thumbnail via AIMUG

Six Talks at AIMUG: Building Agents That Work

Updated:

AIMUG, the AI Middleware Users Group (it started as Austin LangChain), is where I go to see what people are actually building with AI, and to hear what didn’t work. We were talking about agents here long before they went mainstream.

Since May 2025 I’ve given six talks there. AIMUG collected them on my speaker page. Here’s the short version of each, newest first.

TalkAIMUGRecording
Grokbot and Jev, a decision modelOct 2026Watch
LangGraph for fraud detectionJun 2026Premieres Nov 5
OpenClaw: a swarm of local agentsMar 2026Premieres Nov 19
Claude Code’s Chrome extensionJan 2026Premieres Dec 10
Cursor tips and tricksJul 2025Watch
Controlling Odoo ERP over MCPMay 2025Premieres Jan 7

Grokbot and Jev, a decision model

October 5, 2026 · 16 min · Talk page · Watch on YouTube

Two things I’ve been experimenting with:

Takeaway: use code for exact checks, a decision model for bounded judgments, and an LLM for writing and open-ended reasoning. LangGraph lets you combine all three in one workflow.

A slide from my October 2026 talk showing separate LangGraph nodes for code, Jev, an LLM, and tools. One workflow, specialized nodes: code, Jev, an LLM, and tools.

More: slides and demo material · my write-up on Jev and personal agents · recording on LinkedIn

LangGraph for Fraud Detection: Guardrails, Evals and Langfuse

June 2026 · 21 min · Video premieres Thu, Nov 5

AIMUG thumbnail for "LangGraph for Fraud Detection," showing a structured benchmark comparing AI agents. Thumbnail via AIMUG.

How to build a fraud-investigation agent you can trust: LangGraph for the workflow, guardrails around what the agent is allowed to do, evaluations to measure it, and tracing with Langfuse so every tool call is visible.

Takeaway: the agent gathers and explains the evidence; a human and the system of record make the call.

More: my write-up on building an AI fraud investigation agent

OpenClaw: Running a Swarm of Local AI Agents

March 2026 · 15 min · Video premieres Thu, Nov 19

AIMUG thumbnail for "OpenClaw: Running a Swarm of Local AI Agents," reading "Runs on your PC" over a slide titled "What's Actually Connected." Thumbnail via AIMUG.

Running a team of specialized local agents with OpenClaw, and what they’re actually connected to: email, calendar, an Obsidian vault, and more.

Takeaway (straight from the slide): these aren’t API wrappers, they’re real tools. The depth of tool integration is what makes a “second brain” real instead of a metaphor.

More: Building the agentic economy: my second brain powered by OpenClaw · Hermes vs OpenClaw

Claude Code’s Chrome Extension: What Works and What Doesn’t

January 2026 · 16 min · Video premieres Thu, Dec 10

AIMUG thumbnail for "Claude Code in Chrome," showing a web-testing example slide. Thumbnail via AIMUG.

Browser automation with Claude Code driving Chrome, including a web-testing example where the agent opens your local server, fills in a form, and reports what it saw.

Takeaway: an honest look at what works and where it still falls short.

Cursor Tips and Tricks

July 2, 2025 · 13 min · Talk page · Watch on YouTube

A slide from my Cursor talk on installing Cursor on iOS as a home-screen web app. Cursor on your phone: add it to the home screen as a web app. Click to watch.

Connecting Slack and Cursor so I could kick off work from my phone, then watching background agents build, break, and fix a site on their own.

Takeaway: Cursor gives you both options: a quick chat (even through Slack) for background work like checks and unit tests, and the full IDE when you need it.

Controlling Odoo ERP with an AI Agent over MCP

May 2025 · 10 min · Video premieres Thu, Jan 7

AIMUG thumbnail for "Run company with agent," showing the talk on the projector at an AIMUG meetup. Thumbnail via AIMUG.

Connecting an AI agent to Odoo, an open-source ERP, over MCP, so the agent can work inside real business software instead of just talking about it.

Takeaway: with MCP, an agent can act on the systems a business runs on, not just chat about them.

Come learn with us

If you want a glimpse of the next tech wave, come see what this group is working on.

Meetups are free, on the first Monday of each month at 6 p.m. Central, at ACC Rio Grande, RGC 3000, 1218 West Avenue, Austin, with an online option. RSVP for the next meetup, join the Discord, or browse all my talks on my AIMUG speaker page.

Austin LangChain / AIMUG parrot logo


Share this post on:

Next Post
Jev, Calibrated AI Decisions, and Living with Personal Agents: My AIMUG Talk