Agentic Hackers #2 — From Script to SQL Agent

Meetup will take place on Discord: https://discord.gg/QPGKwmuwKD

This time we're going deep on the basics of how agents are built — using a SQL agent as our example. Pre-recorded demos and code snippets the whole way through. We start from a one-liner and build up step by step:

• One-shot script — one LLM call, wrong answer, no verification. We've all been here.
• Tool use — the agent reads real schema, executes real SQL. Things get interesting.
• Structured outputs & the agentic loop — reliable format out, automatic recovery when you don't get it.
• Reflection — making the agent catch its own mistakes before they reach you.
• Context management — a JSON file of known data quirks, metric definitions, and past queries. The model is only as good as what it can see.
• Observability, guardrails & evals — how to actually debug this stuff, trust it, and eventually ship it.

🗣 Open Stage

• Want to present something? The stage is open.
• Want to propose a topic, paper, or news item? Welcome.
• Want to just discuss ideas? Also fine.

Ping me in advance if you want to present so I can update this description.