Paul Tardieu

Ops intern turned "AI guy" | 🇫🇷 🇹🇼 in NYC

[Paul Tardieu @PaulTardieuSp]

Hey X, I’m Paul 👋
- Intern turned “AI guy” at a sports tech startup in NYC
- Our clients: FIFA, NBA teams, Premier League clubs..
- Building the AI system our revenue team runs on
- Posting my agent setup (Claude Code, Hermes, OpenClaw, Grokbot), what works and what breaks

Also: half-French, half-Taiwanese, but my Spanish beats my Chinese 🥴 (sorry mom), used to work in fashion, love photography and (cuban) salsa

Come say hi if you relate to any of the above!

[Paul Tardieu @PaulTardieuSp]

@techguy_x consistency beats timing and luck everytime imo

[Paul Tardieu @PaulTardieuSp]

As an GTM ops that had never written a line of code before march, I asked my (cracked) CTO to help me structure the repo behind the system finding, enriching, scoring and routing our leads.

Here are the files he handed me as references:
- CLAUDE.md: a sticky note telling Claude “your instructions are in the other file.” It’s just a redirect.
- AGENTS.md: the house rules for the AI working on this project. The main one is that the AI can prepare changes but never publishes them live without the repo's owner approval.
- AGENTS.wmill.md: the standard user manual from Windmill, the automation tool, explaining how to build and test workflows correctly. It’s maintained by Windmill, not by us, and gets updated automatically.
- README.md: the guide for humans on how a change goes live. You build and test in a sandbox, a colleague reviews it, and only after approval does it reach the live system our revenue uses.
- Beads: the AI’s to-do list and project logbook. It tracks what’s done, in progress and next, so anyone in the repo can ask “where are we?” and get a factual answer even weeks later, instead of relying on someone’s memory.

[Paul Tardieu @PaulTardieuSp]

@SamVale29 AI won't replace jobs, people mastering AI will

[Paul Tardieu @PaulTardieuSp]

@ben_sage OGs

[Paul Tardieu @PaulTardieuSp]

@bernhardsson @grok can you explain the context or "russian bots"?

[Paul Tardieu @PaulTardieuSp]

@IAmAaronWill I swear
Just started using the app, and it feels so much more TRUE than any other social media
The fact that it's way more based on writing makes people put way more thoughts into what they post
And there are genuinely helpful people and community out here

[Paul Tardieu @PaulTardieuSp]

@alexdanilowicz FUNraising

[Paul Tardieu @PaulTardieuSp]

@StackDhruv exa is good, you can combine it with an open router like https://t.co/qOr2Tpcc9M and it's pretty efficient

[Paul Tardieu @PaulTardieuSp]

@eshaankansal Appying by doing is the new moat

[Paul Tardieu @PaulTardieuSp]

@DenisCerednice1 posting content, AI writing is still so bad

[Paul Tardieu @PaulTardieuSp]

@yongfook this has always been the problem of social media & algorithm
but if you find the right niche with poeple actually trying to grow and share, you will learn so much!!

[Paul Tardieu @PaulTardieuSp]

Check out my hacker house in sf

[Paul Tardieu @PaulTardieuSp]

@lucas_builds maybe this can help? how to make sure your code works is probably the most important
https://t.co/wI6nM2WpLH

[lauren @poteto]

okay now that we're past that tired discussion, let's talk about how we get agents to write more correct code (note: this won't mean the code will be Good, but it should at least work)

this is the entrypoint into your most basic agentic loop: verification. it means giving your agent the ability to verify its own work, most commonly by actually running and interacting with it like a real user would. whether your interface is a cli/api/ui/whatever, you want to start here so that your agents can close the loop. if you want you can even have your agents model the problem in TLA+ and attempt to verify if you have things like state machine bugs.

the point is to give yourself confidence in the agents' work by them producing some kind of artifact that proves that 1) that the work is correct, 2) that it matches what you intended, and 3) is easier to review than reading the whole PR. most commonly this would be a video and/or screenshots for UI work.

everything else like skill packs are really bonus and optional, all you need is a way for your agents to verify their own work to start using them more autonomously. some caveats are that not all domains can be verified easily yet, and that verification is just approximation and can still be wrong. it's no different than trusting that a human actually ran the code they wrote and trusting it

so start here!

[Paul Tardieu @PaulTardieuSp]

@kushbhuwalka "how much did you raise?"

[Paul Tardieu @PaulTardieuSp]

To all my GTME fellows, I built the best inbound engine EVER in only 2 steps:
1. Get invited to #Sculpt2026 by @clay
2. Wear a https://t.co/xJldcpEAdO red mohair sweater

[Paul Tardieu @PaulTardieuSp]

@DenisCerednice1 "Can I check your repo?"

[Paul Tardieu @PaulTardieuSp]

@b4dgerr is prompting still a thing though

[gatsby @0xg4tsby]

This is f*king insane. Google Brain founder Andrew Ng:

"Prompting will die in 6 months.

99% will keep writing prompts.

The 1% will build loops and graphs."

In a 2-hour lecture, he breaks down the path from prompts to agents, loops, and graphs

Prompts → Agents → Loops → Graphs

A prompt gives you an answer

The first 15 minutes are worth more than most $500 agent courses

Bookmark and watch it today

Then read my full guide to loops and graphs below ↓

[Paul Tardieu @PaulTardieuSp]

@shuangling_yin they're too busy doing PR

[Paul Tardieu @PaulTardieuSp]

turning your past life into marketable experiences is not new though. 10 yrs ago success was getting into GAFA, 5 yrs ago into MBB, now it’s getting into YC and exiting. it's the same playbook and the same people, just under a different logo.
what AI actually changed: building is cheaper, and the prospect of an exit promises MUCH MORE wealth x status than those old golden paths.
but scaling isn’t easier.
the filter might not be investors (bluffable), but customers are, and usage even more.
a lot of the AI deals signed right now are FOMO pilots no one opens.
the larpers can exit before the lights come on, but their products won’t.