• punchtime
  • Posts
  • Back to the Future with Troy Osinoff

Back to the Future with Troy Osinoff

Turning Vibes into Ventures

Author

Andrea Puenchera
April 24, 2025

This week I had the chance to sit down with Troy Osinoff, entrepreneur and OG of internet culture. If you’ve ever met Troy, you probably walked away thinking two things:

  1. What a smart, chill dude 🧠 😎

  2. And somehow, he has sense for what’s next 🔮

Troy lives in the future. He doesn’t just see trends before they peak, he creates the energy around them. Whether it’s bitcoin, a meme, or an entire movement, Troy is often there before the masses arrive. And when he shows up, he doesn’t just watch reality unfold, he takes action to shape reality. I always feel that Doc Brown energy around him — all ideas, no brakes, and about to hit 88 mph in his DeLorean.

Just a few of his wild adventures so far:

  • Started creating websites at age 10 when the internet was still dial-up

  • Launched MakeaGif, one of the first platforms turning videos into shareable GIFs

  • Helped bring the last shoe factory in NYC back to life

  • Published author with Penguin

  • Co-founded JUICE, the agency behind campaigns for Paris Hilton, L’Oréal, Nucleus, and more

  • Co-founded Zurp, a Gen-Z credit card where we first crossed paths

  • Backed dozens of startups as a GP at MAGIC Fund — and now through his own fund, yo.vc

📍 You can find Troy on X (Twitter) and Instagram, or at yo.vc where he backs bold founders and experiments with what’s next.

Everything in <60 Seconds

  • In 2017, Troy built an AI tool that auto-hid negative comments — years before most people had heard of LLMs

  • Today, AI is his daily companion — for research, prototyping, and even design reviews straight from ChatGPT

  • He fine-tunes AI to match his style: short, sharp, and actionable (bullet points only, please)

  • He’s blunt: AI will replace most marketers — especially media buyers. Clickbait is easy…

  • His real concern? AI’s cold logic. If the goal is optimization, will human lives be part of the equation?

  • His advice? Don’t overthink it. Just start building, testing, and talking to it like a smart friend.

News, Tools & Resources

Stay ahead of the curve with some news and experimental AI tools.

📰 News from the AI world

  • 🔥 Meta drops A NEW APP called edits
    A new video creation app aimed at streamlining short-form content.

    Get it here on Apple App Store → 

    Read Intro post → 

  • 🏃‍♂️ Humanoids Run a Half-Marathon
    Boston Dynamics meets Nike Run Club — robots just ran 13.1 miles
    Watch the video →

  • 🔍 Washington Post x OpenAI
    OpenAI will now surface Washington Post content directly in ChatGPT search
    Read the announcement →

  • 🧭 Claude’s Moral Compass
    Anthropic opens up about Claude’s thinking and value system
    Explore the AI values drop →

  • 🌊 OpenAI’s Bid for Windsurf
    Why OpenAI chose the rising devtool Windsurf over Cursor — and what that means for the AI IDE wars
    TechCrunch coverage →

🧪 Experimental Tools

  • 🗣️ Distro
    Built by Alex Lieberman — turn any conversation into publish-ready content in minutes

  • 🤖 ChatNode
    An AI-powered customer support tool

  • ✍️ The Good AI
    An AI essay writer that structures and refines your thoughts

  • 💸 Glean
    Smarter accounts payable automation for modern finance teams

🚀 Turning Vibes into Ventures with Troy

⚡️ Early Days of AI: Troy’s First Real-World Use Case

Unlike many who found AI through ChatGPT, Troy had a different kind of “aha” moment — one rooted in practicality and observation.

It was 2017. Troy was at BuzzFeed, watching a co-worker manually moderate negative comments on posts.

Why? Because bad comments tank post performance — which hurt ad revenue.

Most people would think: What a tedious job.

Troy thought: There has to be a better way.

So one of the first things I did when I started Juice with Mike was figure out a way to automate that using AI. I built a tool using Google’s sentiment API that automatically detected and hid negative comments in real time. I called it SourGrapes.

I saw the huge potential of AI. It was already taking over small, repetitive jobs.

That’s his AI origin story. Not theoretical. Not academic. Just pure, hands-on problem solving with early tools.

SourGrapes still lives on today at sourgrapes.ai, but Troy envisions something more:

I want to turn it into a whole brand suite. A reputation management tool.

📊 Practical AI: Troy’s Daily Use Cases

Fast forward to now — Troy’s AI stack is part of his daily workflow. When asked about how often he uses AI these days he said:

Constantly. I use it for everything. Instead of Googling, I just ask ChatGPT. Whether I’m researching stuff for my dog, my health, or building tools — I’m using AI.

What’s really cool is how he personalizes responses to his needs and conducts research

I train GPT to be more useful to me specifically. Be brief. Ask more questions. Bullet point everything so I can read at a glance.

When asked about what aspect of AI saves the most time or brings the most value, Troy honed in on research.

Definitely research. Instead of asking people or hunting for sources, I can just get direct answers. The time saved is insane. You can ask a question, get context, dig deeper. It just makes you a better researcher. The barrier to learning is way lower now.

He even used GPT to get feedback on what he’s already built. He drops his Figma mockups into ChatGPT just to see what it says.

When I was building an app recently, I literally took Figma screenshots of the onboarding flow and uploaded them to GPT. I asked, “What do you think of this onboarding?” And it said, “This isn’t onboarding, Troy. These are just settings.” So I redid the entire flow based on that feedback. It’s wild how useful it is as a second opinion.

🧠 What’s Left for Humans?

But not everything about AI is playful and productive. I asked Troy about the biggest danger AI poses, and let me tell you, he didn’t sugarcoat his answer. Hello Skynet scenario and T-800.

I was talking to my grandma yesterday about this topic and I said it could kill the entire population of the planet. She asked, “Why would AI kill humans?” I said—logically— if you had AI that's powerful enough, and integrated in everything we do, and the government, or someone asked, what's the most effective way to make the planet healthier, more sustainable and effective. The obvious answer to AI would be destroy all humans.

He emphasizes that it’s not about malice, but logic:

We don't really act good for the planet. We don't act as good for ourselves. We’re emotional, we’re not logical creatures. We destroy the planet. Robots don’t. If an AI is tasked with optimizing society… humans might not be part of that equation. We require a lot of upkeep. Why would AI want us?

He's not trying to be dramatic, he’s just trying to think rationally.

I’m not a doomsday prepper. It just is the most rational answer to me.

📈 Rethinking Marketing and Content in an AI World

When it comes to marketing, Troy doesn’t hold back.

I think AI is going to replace most marketers — especially media buyers.

Why?

The platforms will know the best strategies. They already have the data. It’s all optimized for performance, and AI will know way better and faster which copy converts.

Humans are emotional and predictable. There’s so much data. Why do you need a human to come up with ideas anymore?

He laughs and says:

Clickbait is easy…

But it’s not just marketers… Any creative field driven by repeatable human patterns is vulnerable to the same ruthless optimization. In a world where content is formulaic and data-backed (see TikTok), AI is always a step ahead.

I also play around with Suno, it’s for AI-generated music. I think music is going to be huge. It’s already algorithmic — we know the formulas that work with audiences. AI will just enhance that even more.

🎵 Troy’s AI Stack & Advice

Troy’s AI stack:

  • ChatGPT – Conversational reasoning, general queries

  • Perplexity – Research with citations

  • Cursor – Fast prototyping, code generation

  • Replit – To build things without being super technical

  • Claude (Anthropic) – Sometimes better for creative

  • Suno AI - Generate music based on moods and themes

His advice to beginners?

Don’t be intimidated. There’s no dumb question. Just start talking to it like you would talk to a smart friend.

True to his doer mindset, Troy’s advice is simple: just start playing around. Build something. Break something. Drop your ideas, apps, or challenges into GPT and ask, “What do you think? How can I automate this?”

The magic is in the back-and-forth. You learn more by trying than waiting to become an expert first.

Final Thoughts

I always leave conversations with Troy feeling energized. Hearing how he’s thinking, building, and spotting trends before they go mainstream — it’s a reminder of how much perspective help me make progress myself.

Different angles. Real stories. Cool experiments.

Until next time — keep building, keep exploring, and rock on.

-Andrea

What did you think of today's edition?

Please reach out and share what I could improve

Login or Subscribe to participate in polls.