GPT-5 Pro sets a new record

Edition 128 | October 16, 2025

If the CIA isn’t working on this right now, I want my tax dollars back.

Welcome back to Building AI Agents, your biweekly guide to everything new in the field of agentic AI!

In today’s issue…

• AI coworkers are here

• A free interface to over 50 MCP servers

• 🎓 Reminder: Join our free webinar on OpenAI’s Agent Builder today @ 4pm CT

• How to run a whole virtual coding team

• GPT-5 Pro breaks a general intelligence record

…and more

🔥 IN CASE YOU MISSED IT

Readers’ favorite items from the past week

1. OpenAI’s recap of everything it launched on DevDay

2. The 2025 State of AI report

3. McKinsey’s “agent-first business” playbook for CEOs

4. n8n raises $180 million for its low-code agent builder

5. Tasks agents can do nearly 100% on their own

 🤖 AGENT OF THE WEEK

Welcome back to Agent of the Week:

Normally, your agent has to hit Google Sheets or Airtable every time it needs data.
Now? n8n Tables store everything inside n8n — fewer API calls, faster runs, cleaner builds.

So to test this, I built a CRM Sales Assistant Agent — it keeps client and deal data in n8n Tables, then lets me chat to pull summaries, list won deals, and even sum revenue by month — answering all my sales analytics questions.

One agent, one native CRM table. Simple, but seriously powerful for everyday client questions.

How I Built It:

1. Open n8n: Create a new workflow (free plan is enough)

2. Create a Data Table: leads_example with columns: name, company, source, lead_stage, assigned, email, deal_value (number), contact_date (date), status (won/lost/pending)

3. Import data: Automatic Trigger → Google Sheets: Get Rows → Data Tables: Insert Rows (you’ll see “11 out → 11 in”)

4. Agent: Chat Trigger → Router (name / status / date) → Connect Data Tables: Query → Calculator to sum money → AI Agent → answers things like “Paid deals in September” or “Total from Orion Motors”

5. Add Guardrails in Prompt: “Use only table data, never invent facts. Show a one-line math breakdown when summing. Output 1–2 sentences + a mini table.”

6. Ask questions to the CRM Sales Assistant Agent: “How many paid deals did I have in September?” or “tell me about all projects we did for Orion Motors this year” or “How many paid projects did we close last month”

Bonus: Want the detailed video walkthrough and the exact n8n agent template? Get access to both in our Building AI Agents Community here & check out the full tutorial here!

Till next week,
AP

P.S. What agent would you build? Reply with your idea and I’ll build it next week!

If you find Building AI Agents valuable, forward this email to a friend or colleague!

‼️ ESSENTIAL INTERVIEW

AI coworkers are here

A TIME interview with Amjad Masad, CEO of Replit, in which he discusses the rise of agentic AI coworkers, which can increasingly operate semi-autonomously for hours on end. These agents are actually improving even faster than suggested by the report I featured on Monday, he argues, and gives advice on how to stay ahead of the curve as they increasingly disrupt entry-level jobs.

Masad was a key figure in setting off the agentic coding boom by creating Replit agent, which is now one of the most popular AI-powered coding tools and has brought the company’s revenue from $2.8 million to $150 million in less than a year.

 🎓 LIVE WORKSHOP: OPENAI’S AGENTKIT

OpenAI just launched their no-code AI agent builder AgentKit! We’re going live Today @ 4PM CT to show you how to start using AgentKit, how it compares to other tools like Zapier & n8n, and simple agents you can build today with it!

RSVP here!