The founders and developers who built Topic, Chiba and Nikhil Aitharaju joined us to chat about content workflows, production, finding new content ideas, AI and content, and more in this AMA. For more on their product, check out the links below:
If you want to try it out, use this link to get 30% off: usetopic.com/superpath
Ryo: There are lots of new AI tools out there, your community is likely familiar with a few of them already. And for anyone who’s tried them out, it’s clear that the text-generation capabilities of AI are definitely unlike what was available in the past. However, we see it not as something that’s going to steal people’s jobs, but over time it will change them. In its current and future states, AI will still need a driver, a person who has an understanding of the goals, narrative, and brand that the content serves. By making it easier to publish certain types of content, it’s going to raise the bar, requiring everyone to find new ways to differentiate.
In terms of how we use it in our own product, we use GPT-3 for a variety of use cases, including generating content ideas, writing descriptions, semantically clustering topics. GPT-3 is really good at tasks like generating short paragraphs of text (tweets/titles/descriptions/emails), classifying text, summarizing text & rephrasing text. A major shortcoming is you can't have GPT-3 generate an entire high quality long form article ( >1000 words) by providing it with just a title / synopsis. You need to pair it with a human to get a high quality output. So we built our platform with the human-in-the-loop approach that starts with a platform of curated research, unlike other GPT-3 long form text generation tools. We plan on adding more functionality so that you can build off of an outline built in Topic and use AI to quickly flesh out a first draft.
Ryo: Taking a step back from the specific content marketing use case, the truly unique thing about GPT-3 as a technology is how flexible it is in terms of the problems that an average developer can use it for. Typically, it would take a team of data scientists to train a model to successfully handle a single use case, which makes it cost prohibitive for most companies. GPT-3 basically allows you to skip that step and simply give it a few examples to emulate.
What this means is that you see GPT-3 popping up in all sorts of places, from writing code in Github Copilot to brainstorming blog post ideas. So GPT-3 has made it easy for developers to tell the AI what to do, the next step is building products that make it easy for end-users to tell the AI what to do. Right now it’s still in the early stages of product development, where there isn’t enough tooling yet to for end users to efficiently describe what their end product should look like. So we see it being used most effectively for simpler use cases like brainstorming ideas for a title. This is what it is best for right now. Simpler use cases where you can maintain tight control over the output.But as the tooling develops and more sophistication is added to products to help steer GPT-3, I anticipate that we’ll see it being used in more complex situations, for example, learning a writer’s tone of voice and incorporating that into paragraphs that it generates. Or being able to continue a cohesive narrative without getting sidetracked or repeating itself.
Nikhil: Our product is designed to help editors and agencies speed up content production process without compromising on quality. We don't target writers or bloggers at the moment.
Nikhil: That's correct. Usually they are the ones building content briefs and making final edits before the post goes live. We'd love to target writers/ bloggers at some point in the future by splitting our product. However at our current price point and the way the product is built, it makes it a little difficult at the moment.
Nikhil: That's a good question. In my previous message, when I said writers, I was referring to freelance writers specifically. For your specific use case, if you're responsible for both the research and writing part, I would recommend the product.
Ryo: To be honest, the majority of developer-type content that I consume on a daily basis has been to solve a specific problem and the editorial aspect is a secondary concern. What most developers care about is finding a solution that’s up-to-date, clear, and concise. In terms of trends, I can only speak from my own experience, but I’ve seen the quality of what’s out there significantly improve as there is more and more incentive to create developer-literate content, one example that comes to mind is https://blog.logrocket.com/, they’ve done a good job of creating a content machine that’s constantly publishing new niche dev content. Sites like https://alistapart.com/ have been a mainstay for more editorial dev content and I enjoy reading that too, but I’ve found that it comes up less among our customer base since typically more editorial sites are creating a smaller number of carefully crafted articles talking about the latest trends rather than targeting the masses through search.
Nikhil: They have been really helpful in building brand awareness and domain authority. They do bring in leads however it isn't the biggest driver of our leads. We've one in the works to let you grade your content for free.
Nikhil: It's been word of mouth for us. Luckily communities like Superpath help spread the word for us.
Ryo: Luckily, since this was the second time Nik and I have started a company together, we were already prepared for the #1 lesson, which is that getting a project off the ground typically takes much longer than initially anticipated. We initially thought it would take us a year, it took us two.Second biggest lesson while bootstrapping Topic has been how important it is to be able to admit defeat and scrap your own work to start fresh, and to be able to distance yourself from your own work and put ego aside. That first year we spent building an SEO A/B testing tool that we eventually stopped working on because of a lack of traction. There is an optimal time to quit and switch gears and getting that calibrated correctly is really important for any creator.
Nikhil: One of the things we started realizing recently is not having a third co-founder. Ryo & I love building product, marketing/ sales is something we do for growth, however it isn't our strongest suite. Next time we'd love to find a third co-founder with complementary skill sets.
Nikhil: We talk to our customers every day which helps us get up to speed with the latest SEO and content news. Also, we're responsible for growing the business so that acts as a forcing function for us to stay up to date with latest trends and marketing strategies. My go to source for learning is usually podcasts and blogs. Here are some of my favorite blogs: Superpath, Animalz, Grow and Convert, Backlinko & Swipefiles.
Ryo: Yes, we are account managers and customer support so that definitely helps. But we’ve also designed the product to make it easier for customers to start conversations with us. For example, we have a custom “Chat with Us” button at the top of the app. If you hover over it, a box with our faces on it encourages users to talk to us (definitely gets clicked on more than a generic intercom bubble). And when we’re rolling out a larger feature, we’ll gate it by embedding a CTA within the product that links to a Paperform for people to schedule a time. When it comes to connecting with customers, there’s a lot that can be baked into the product.
Nikhil: For us, it's been word of mouth. We do lots of random little things that help spread the awareness about our brand. I can't attribute to one channel for our growth at the moment.
Nikhil: We recently generated a report , it said those pages aren't generating any leads. I find it hard to believe, may be attribution is broken. The reason we built those pages is to help the buyer make an informed decision. Lead generation wasn't the main goal
Nikhil: Ahrefs, Keywords Everywhere, Search Console are couple of my