14.ai Dispatch #17
Password resets in 15 BC???
MCP servers, Monopoly with crypto bros, and YC founders laying on yoga mats in SoMa.
How long would you wait on hold?
You know the feeling.
You’re desperate to fix something. You call support.
Then it begins.
The hold music. A four-bar classical loop that slowly eats your soul.
Ask yourself: how long would you wait?
Event recap: AI agents for fintech night


Crypto, Monopoly, and pita bread. That’s a first.
Between rounds of Monopoly Deal and debating how to stop your AI agent from “accidentally” wiring $5M, we finally landed on the real topic: building reliable fintech + crypto agents.
Thanks to everyone who showed up and kept the competition unhinged at the table and in the code.
Founders & Engineers’ Typing Competition at #SFTechWeek
Think you’ve got the fastest fingers in the Bay?
Join us at the 14.ai HQ for a casual but EXTREMELY SERIOUS typing showdown as part of the 2025 SF Tech Week.
Startup leaders and engineers put speed, accuracy, and stamina to the test. Bragging rights are on the line.
Connect with fellow techie and prove your keyboard game is as sharp as your pitch.
Building MCPs you can trust
“Security” usually means someone bought Okta and called it a day.
Not here. Wojciech Błaszak (Golf, YC X25) pulled up and reminded us that MCP servers in production need:
Audit trails (aka receipts for when your intern breaks prod)
Proper auth (no, “password123” doesn’t count)
Prompt-injection defense (because users are gremlins)
Big thanks, Wojciech!
Next week’s workshop: Building knowledge bases that grow with your product
As your product evolves, your knowledge base has to keep up. Join us for a live discussion with Nivedha Venkatesh (CEO @ Pageloop), on what it takes to scale AI knowledge bases beyond the basics. We’ll explore how fast-moving teams can balance speed with accuracy, handle multi-product complexity, and design knowledge layers that evolve instead of break.
15 BC Batch
YC in 15 BC was insane. Caesar raised a $2M seed round from the Senate and still couldn’t get product-market fit.
In case you missed it
Sleepovers: candidate trials at 14.ai = Barry’s Bootcamp, communal meals, shipping code in pajamas.
Prompt pokes: 30 people tried to jailbreak our system prompt in five languages.
Routine fix: breakthroughs live in bedtimes and Barry’s, not Bali. Sorry Bali.
Claude code: Effect + TypeScript = AI that writes clean, production-grade code.
Chat support: book time with us before we automate ourselves.
Cool jobs at 14.ai (and friends)
🛠️ Founding Engineer @ 14.ai (San Francisco)
⚙️ Director of Customer Support @ Shippo (Remote)
💻 Customer Success Manager @ Storylane (US, India, Remote)
What we’re reading & listening to this week
Marie: Founder Mode: Chris Best, Founder & CEO, Substack, The Social Radars
Matthew: Why React Re-Renders, Josh W. Comeau
Maya: AI isn’t replacing radiologists, Works in Progress and Deena Mousa
Michael: Tau² Benchmark: How a Prompt Rewrite Boosted GPT-5-mini by 22%, Przemysław Hejman
Ugo: Interacting with Thoughtful AI, Liu et al., 2025
🫡 Until next time. May your prompts stay secure, your cards deal fair, and your typing competitions go smoothly.











