Beyond the slop cannon

An atmospheric physicist posts a careful thread about uncertainty bands in Arctic ice models. Twelve likes.

Some dude with a Roman statue avatar posts "CLIMATE CHANGE IS A HOAX." Ten thousand retweets.

The platforms know what they're doing. Engagement equals revenue, outrage drives engagement, therefore outrage drives revenue. They dress it up with words like "meaningful social interactions" but we know what really moves the needle.

There’s a version of the future where AI makes this worse - a diabolically optimized slop cannon firing at the human face, forever - but I think there are some reasons to be optimistic.

Not by detecting misinformation. By being a better listener than any of us.

I've been testing new AI products for the fund. Started skeptical, especially after the ChatGPT wrapper gold rush of 2023. But something's different about this wave.

Take Boardy. Twenty minutes on the phone with an AI about what I actually do, what kind of people give me energy, what conversations I want to have. I found myself being honest - telling it what I believe and why.

The AI didn't judge. It asked follow-ups. When I contradicted myself, it gently probed instead of calling me out. A week later it introduced me to someone, and it was an interesting conversation. Next time it reaches out, I’ll take the call.

Known throws out the swipe in favor of a conversation with an AI matchmaker. As soon as I started the call, it dove right into asking me about where I grew up and how that shaped me as a person - the antithesis of what shows up on a Tinder bio.

Superposition (disclosure: F4 Fund portfolio company) is doing this for technical recruiting. Instead of those insane job requisitions that demand 10 years experience in frameworks that have existed for 3, founders just talk about their company. The AI figures out what kind of person would thrive there and goes out to find them.

What makes this work? Infinite algorithmic patience.

Humans are terrible listeners. We wait for our turn to talk, pattern match to our own experience, finish people's sentences. My wife has pointed this out roughly one thousand times. She's right every time.

AI doesn't get tired. Doesn't need to be interesting. Has no ego to defend.

This all sounds great, but listening is the easy part.

What matters is what comes after - the mysterious alchemy that separates great human networkers from the people who just introduce everyone they know to everyone else they know.

I know a headhunter who's placed half the C-suite in gaming. She doesn't just listen to what companies say they want. She reads between the lines and picks up on the stuff they don't know to ask for.

Can AI do this? Can it actually develop taste and judgment? The jury is still out.

A great matchmaker once told me she ignores what men say they want in a partner and listens to how they describe their exes. The woman they describe with the most energy - even if it's negative energy - tells you their actual type. That's the kind of pattern recognition that matters.

These new systems are betting they can find those patterns. Not just "you both like hiking" but "you both get animated discussing the same kinds of problems" or "you have complementary blind spots" or whatever subtle signal actually predicts connection.

Traditional platforms are attention routers - moving eyeballs around, maximizing surface area. These companies are trying to build understanding engines. Whether they succeed depends on if AI can develop something resembling taste.

Will AI become as good as the best human matchmakers? Probably not. But it might not need to be. It just needs to be better than swiping through photos or broadcasting into the void. That's a pretty low bar.

The question isn't whether machines can listen better than we do - they already can. It's whether they can develop the judgment to know what to do with what they hear.

If they can, we might finally have an alternative to the slop cannon. If they can't, at least we'll have had a few decent conversations along the way.