Gabor Cselle on LinkedIn: VidCon takeaways so far: 1. Gotta niche down. Decide who you want to be… (2024)

Ran into Beau Hartshorne at VidCon - we caught up on all the things that happened since our time at Y Combinator in the 2000s. He’s now launched Strobe which is a platform to evaluate creators at scale.

14

Caught up with Sasakthi, Buddhika and Olya from the YouTube Dubbing team today at VidCon - we incubated their product in Area 120 in what feels like a million years ago, today they were on the big stage at the conference!

48

3 Comments

Jenny Hoyos at VidCon “How to make anything go viral.” (She has >1B views on YouYube.) Takeaways:1. You can make a video about any topic, but the hook and the idea have to be great. This process she calls “Turn coal into diamonds.”2. What makes a good hook?Visually shocking. E.g. Her holding a hammer breaking a home wall, a McDonald’s sign on fire.Establishes the right expectation, e.g. “Mom, I’ll buy anything you can carry.” Image of Jenny carrying a lot of stuff.Establish the promise and then immediately start delivering - e.g. her Mom started buying immediately.Foreshadowing - there needs to be a reason to watch through the end, there is a question that is answered. There needs to be a goal by the end of the video. E.g. “I’m surprising my family with their favorite lost things. They don’t know these are the things are stole from them.” Now you need to watch to see how the family reacts.Think: What words can we use for the viewer to immediately get it?- “Who has the best dollar menu?”- “How much ice cream can you get for a dollar?”- “Is it faster to go inside or the drive through?” Everyone knows what “fast” is, McDonald’s is, a drive through is.3. For any new video, all she starts with is the hook and the second line, e.g.- Hook: “Is it faster to go inside or the drive through?”- Second line: “… because my mom is already driving through and I’m running inside”The rest of the video is always acted in real time, unscripted. She films more video than she needs.4. Personal brand is important. Her lens is always “how can I be a cheapskate?” For her first 20 videos she tried a bunch of stuff, eg crypto, forex, crypto mining, and then broke out with the cheapskate topic, got to 10M views at about the 20th video. The cheapskate angle was a blue ocean. Few people want to portray themselves as a cheapskate online.Character stays consistent across all the videos.5. Very important to keep a high energy vocal tone throughout the video to keep people retained to watch to the end.6. Per thousand views on long form, she makes $5. Per thousand views on short form, she makes $0.12. This is with a 40% US audience.

11

Getting some initial user feedback from Samuel Odio for something I’ve been building at South Park Commons. Always a big moment when you first see your product through someone else’s eyes.

50

3 Comments

Met with a friend today who just left an exec role at Google. He’s thinking about starting something. He wrote down some of my advice so wanted to share in case it’s helpful:1. Give yourself time to do -1 to 0. Don’t jump into 0 to 1 just because investors are ready to give you money. Nothing worse than spending years of your life doing something you’re not actually that excited about, but have committed to get to an exit.2. Be clear about what you want, and honest at a minimum to yourself. Do you want to be a famous startup founder? A $1B exit? Or a $2M / year lifestyle business? Invent groundbreaking technology? Be helpful to a particular combination of people?3. If you just want something that make $2M / year, watch all the Starter Story videos on YT.4. Build with the latest generation of tooling, especially it’s been a while since you last built. Build on Vercel or Render and Supabase, not EC2 and RDS. Build apps in FlutterFlow, native only makes sense in limited cases.5. There’s now a deep pool of global talent that is going to be a lot less expensive than SF. It requires more filtering and more management.

242

13 Comments

Excited to share that I've joined the South Park Commons community to explore new ideas and meet fellow builders!

83

14 Comments

Congrats Christopher Pedregal on today's launch!

7

Do you ever type a question into ChatGPT, find the answer lacking, and then copy-paste the question into Gemini or Claude to see if their answer is better?

This content isn’t available here

Access this content and more in the LinkedIn app

4

2 Comments

I've tried most of the text-to-UI-design and text-to-frontend-code generators (trace.zip, usegalileo.ai, v0.dev, ...).This space is where GPT-2 was a few years ago: Promising, but not yet useful beyond demos.What will the unlock be? 2 guesses:a. MOAR training data - these models so far are only able to write code and UIs in very specific styles. I'm guessing training them requires a ton of data prep in getting inputs and outputs into the right format, and that takes time.b. Next-gen foundation models like GPT-5. @sama has been hinting at how capable GPT-5 is for months (most recently on 20VC), and maybe all we need is a better LLM that can reason about UIs more effectively, because it has seen all of them, and all relevant patterns.

28

1 Comment