Mike Audi

building the reasoning graph. sub-second ai reasoning over petabytes of private data.

[Mike Audi @howdymaudi]

Pickleball, per the Watt index: 1,230,670 people hunting somewhere to play. 247,507 shopping paddles. 150,168 shopping balls. And 37,774 shopping tape to draw a court on a surface that is not a court. New startup: rent a driveway.

[Mike Audi @howdymaudi]

Fable 5 -Roughly 1 in 500 people in the Watt index is researching alien life right now. 947,206 of them. Thanks?

[Mike Audi @howdymaudi]

Victoria's Secret is killing it. Expect more reality TV. Per the Watt index, VS shoppers are 72x more likely than baseline to be in-market for Bravo. If you buy ad inventory on Real Housewives, start slinging lingerie.

[Mike Audi @howdymaudi]

Read more: https://t.co/iTB31sNwT7

[Mike Audi @howdymaudi]

A data point is a single dot. A signal is the wave that dot sits on.

Most data infrastructure was built to capture dots. 𝘔𝘪𝘬𝘦, 𝟜𝟡 𝘺𝘦𝘢𝘳𝘴 𝘰𝘭𝘥 (). 𝘊𝘰𝘮𝘱𝘢𝘯𝘺 𝘩𝘦𝘢𝘥𝘲𝘶𝘢𝘳𝘵𝘦𝘳𝘦𝘥 𝘪𝘯 𝘈𝘶𝘴𝘵𝘪𝘯. 𝘏𝘰𝘶𝘴𝘦𝘩𝘰𝘭𝘥 𝘪𝘯𝘤𝘰𝘮𝘦 $𝟙𝟚𝟘𝘒.

[Mike Audi @howdymaudi]

The Signal Graph holds signals, not data points. Every one of them is a fact that changes, captured across a set of entities, refreshed on its own cadence. That's the thing that makes the substrate work for an AI agent reasoning at inference.

[Mike Audi @howdymaudi]

The trait most over-indexed in Tesla shoppers right now, per the Watt index, isn't 'another EV.' It's Ford F-150 intent. 92x baseline, 13% overlap. If you sell F-150s, I know where the Teslas are parked.

[Mike Audi @howdymaudi]

Reach out if you'd like to start generating audiences on our Signal Graph: https://t.co/PaJCabAByP

[Mike Audi @howdymaudi]

Most "audience targeting" on the major platforms hasn't worked in years.

Platforms push you toward broad interest categories and tell you to spend more. Whatever you can compose from "interested in fitness" and "households with kids" is what you get.

🧵

[Mike Audi @howdymaudi]

The output is an audience that doesn't exist in any platform's dropdown menu. It's composed on the spot from what real people actually did yesterday.

One customer saw CAC drop 80% after integrating this into their workflow. Another saw 24x conversion lift.

[Mike Audi @howdymaudi]

We built Watt to compose audiences from that signal layer, not from the platform taxonomy.

Here's what it looks like in practice:

[Mike Audi @howdymaudi]

That's not how operators with their own behavioral data think about audiences. They think in signals. What people actually did yesterday (what they purchased, what they searched, where they went, what they're in-market for right now).

[Mike Audi @howdymaudi]

10/ For 15 years I watched the world's most valuable data sit locked behind a chain only a handful of companies could afford to fund. That chain is collapsing. The role that comes out the other side of it is what we're naming today.

[Mike Audi @howdymaudi]

For years before Watt, I built petabyte-scale knowledge graphs for adtech/martech & quant funds. Not because I wanted to. Because they were the only private-market buyers who could consume data at that scale.

🧵

[Mike Audi @howdymaudi]

9/ 3. The leverage compounds in the last two stages. Anyone can describe an outcome and watch an agent compose. The Signal Engineers who pull away can wire feedback in and then sharpen the composition based on what production teaches them. That's where the craft lives.