Build,
Code
& Ship

CUNY AI Journalism Lab: Builders  ·  Class 6  ·  April 9, 2026
BUILD, CODE & SHIPCUNY BUILDERS — 2026

Today

  • 0:00   Show & tell
  • 0:15   Codex writes code (app + VS Code)
  • 0:35   MCP: giving your assistant new powers
  • 1:00   Subagents: parallel autonomous work
  • 1:25   Your capstone + discussion
  • 1:45   Wrap-up
BUILD, CODE & SHIPCUNY BUILDERS — 2026

Show
& Tell

What did you use your assistant for this week?

What worked? What was missing?

Did you improve your AGENTS.md?

BUILD, CODE & SHIPCUNY BUILDERS — 2026

Codex
Writes
Code

Two ways to use Codex for coding:

Codex App

Upload files, run code in a sandbox, schedule tasks. Quick and self-contained.

VS Code Extension

Sees your whole workspace: AGENTS.md, context files, source lists. Builds where your context lives.

BUILD, CODE & SHIPCUNY BUILDERS — 2026
Build together

Beat Monitor
Script

"Look at my context/beat-notes.md and build a Python script that checks the RSS feeds listed there for new articles in the last 24 hours, filters for articles mentioning my key topics, and saves a summary to daily-report.md"
watch it write run together iterate
BUILD, CODE & SHIPCUNY BUILDERS — 2026

MCP
Servers

Your assistant
isn't limited to
what's in your
folder.

MCP (Model Context Protocol) = a standard for connecting AI assistants to external tools. Think of it as apps for your assistant.

Without MCP
Your folder
With MCP
The world
Today: install one MCP server together and watch it work.
BUILD, CODE & SHIPCUNY BUILDERS — 2026

A Few Examples

MCP Server What it does Newsroom use case
Chrome DevTools Navigate, click, screenshot, inspect DOM & network Capture & inspect how a page loads, debug a prototype
Playwright Automate browsers via accessibility tree Collect data from public websites on a schedule
YouTube Transcript Pull transcripts from any YouTube video Get full text of a press conference or hearing
NotebookLM Query your notebooks, get citation-backed answers Research a document trove with grounded citations
WhatsApp Read, search, send messages & media Send story alerts to a group, search past conversations
Google Drive Search, read, create files; Docs → markdown, Sheets → CSV Pull a shared doc, read a spreadsheet, update a brief
UNHCR ← today Query refugee stats by country, year, demographics Get displacement data for a story without manual download
BUILD, CODE & SHIPCUNY BUILDERS — 2026
Live install

UNHCR
MCP

Connect your assistant to live UN refugee data. One install. Then ask it real questions about displacement worldwide.

github.com/rvibek/mcp_unhcr
install configure query live data
BUILD, CODE & SHIPCUNY BUILDERS — 2026

What just
happened

Prompt

You described the goal

MCP Tool

AI called UNHCR API

Result

Saved to your workspace

This is what "agentic" means: it reasoned, used a tool, took an action. You didn't touch a browser.
BUILD, CODE & SHIPCUNY BUILDERS — 2026

Sub-
agents

Codex can now spawn specialized agents that work in parallel and consolidate results, without you directing each step.

BUILD, CODE & SHIPCUNY BUILDERS — 2026

How They
Work

One orchestrator.
Many workers.
Fresh context each.

Three built-in agents ship with Codex:

explorer

Read-heavy analysis: maps documents, traces code paths, researches background

worker

Execution-focused: implements tasks, processes batches, writes outputs

default

General-purpose fallback for anything not assigned to a specialist

You can also build your own →

BUILD, CODE & SHIPCUNY BUILDERS — 2026
Custom agents

Build
Your
Own

One TOML file.
One specialist.

Drop a .toml file in .codex/agents/

name = "fact-checker" description = "Verify claims against public records" model = "gpt-5.4-mini" sandbox_mode = "workspace-write" developer_instructions = """ You are a fact-checking specialist. For each claim, find the primary source. Flag confidence level: confirmed / unverified / disputed. Always include the source URL. """ [mcp_servers.search] url = "https://search-api.example.com/mcp"
BUILD, CODE & SHIPCUNY BUILDERS — 2026

Subagents for
Journalism

Source investigation

"Investigate this company. Have explorer map all public filings, worker extract financials, and default write the brief."

Batch research

Give it a list of 5 sources. One worker per source, all running in parallel. Results consolidated into one report.

Story verification

"Check this draft. Have explorer verify every factual claim and worker find conflicting sources. Flag anything uncertain."

Coverage monitoring

One agent per publication, running in parallel. "How are these 10 outlets covering this story differently?"

Up to 6 parallel threads by default  ·  Each agent gets a fresh context window

BUILD, CODE & SHIPCUNY BUILDERS — 2026
Try it live

Parallel
Research

"Research [topic on your beat]. Spin up subagents to map the key players, find recent developments and data points, and write a 300-word brief with story angles."
you describe the goal codex picks the agents one consolidated output
BUILD, CODE & SHIPCUNY BUILDERS — 2026

The Full
Stack

MCPSkillsSubagents
KitchenIngredientsRecipeSous-chefs
WhatExternal data & toolsReusable instructionsParallel autonomous agents
DifficultyMedium (install)Easy (markdown)Medium (prompt pattern)
Use caseSearch, archives, CMSInterview prep, digestBatch research, investigation
BUILD, CODE & SHIPCUNY BUILDERS — 2026
Discussion

Your
Capstone

Where is your capstone project right now?

Which layer helps yours — MCP, skills, or subagents?

What external data or tools would make it stronger?

BUILD, CODE & SHIPCUNY BUILDERS — 2026

Human
in the
Loop

Never publish directly

Every AI output needs editorial review before it goes live

Cite sources with links

If it can't show you where it found something, be skeptical

Flag uncertainty

Ask it to say "I'm not sure" instead of filling gaps with plausible-sounding text

Keep sensitive work offline

Source identities, embargoed stories, off-the-record material — never in AI tools

BUILD, CODE & SHIPCUNY BUILDERS — 2026

The 30-Day
Challenge

Use your assistant daily for 30 days.
Add one thing per week — a file, a skill, a tool.
Note what works. Share wins with your team.

BUILD, CODE & SHIPCUNY BUILDERS — 2026

What You
Built

01

Personal AI assistant

AGENTS.md + context files + skills that know your beat

02

Interview prep skill

A reusable workflow you'll use before every interview

03

Beat monitor script

Codex-built Python tool — without writing Python

04

MCP connection

Your assistant now talks to the world outside your folder

05

Subagents

Parallel research with named specialists — just shipped in Codex

Thank
You

flo@mizal.ai
li/x: @fdaudens