• Brainstorms
  • Posts
  • Brainstorms Issue #14: GPT-3 Tool for Twitter

Brainstorms Issue #14: GPT-3 Tool for Twitter

Brainstorms Issue #14: A Twitter Tool I'd Pay For

January 28, 2021

Brainstorms Issue #14

The Idea: GPT-3 (explained below) Tool for Twitter

“Twitter feels like LinkedIn but for cool kids. Even with my small following, I’ve seen a lot of success in making business connections. For this reason, I want to keep leveraging Twitter to grow my business. The problem is, I struggle to come up with compelling Tweets that are interesting and relevant to my audience.”

- Twitter User

The Opportunity

Someone should create a GPT-3 tool for Twitter. 

GPT-3 is an artificial intelligence tool that can answer questions, create designs and code, and can even write poetry from simple strings of text. Not to sound too much like a fan-boy, but the GPT-3 tools I’ve used blew my mind. If you haven't checked it out, do it. You won't be disappointed.

The idea is simple, create a tool leveraging this technology that helps people come up with tweets relevant to their domain or audience’s interests. I see it working like this:

  1. Feed the tool a list of your most popular tweets, accounts that you follow, and accounts that follow you.

  2. The tool uses this information to create tweets that you can either use directly or edit slightly.

This would eliminate a lot of work for coming up with relevant and punchy tweets for your audience.

Based on google search, there seems to be demand for a tool like this. Specifically, “tweet generator” gets about 3,000 monthly queries.  

Market Background & Opportunity Size

GPT-3 Background:

GPT-3 is a natural language processing tool that uses deep-learning to create human-like text.

GPT-3 was created by Open AI, a company founded by Elon Musk and Sam Altman, and was released in beta in June of 2020. While it’s still new, many thought leaders have expressed that GPT-3 is the largest leap we’ve seen in artificial intelligence in many years.

It's currently in closed beta, but some really impressive tools have been created using the technology. Check out, Copy.ai, Dover, or AI Dungeon to get a better understand of how the technology works.

Why Build on Twitter?

Contrary to what some might say, Twitter has been making a comeback and is still an incredibly popular platform. Here are a few interesting stats on why building a tool for Twitter users makes sense:

  • Large audience: Twitter has more than 150M daily active users.

  • Affluent audience: 77% of Americans who earn $75,000 or more use Twitter (source linked below).

  • Motivated audience: Twitter is where creators and makers go to build an audience. The use of Twitter is seen as an investment in their network rather than just a place to update. This is evident by Twitter’s acquisition of Revue, a newsletter publishing platform (this happened last Tuesday).

Note: If you’re interested in more stats on Twitter, look at this.

The Pain Points

There are a few factors that go into growing your Twitter audience. The two primary concerns are both the quantity and quality of tweets. Knowing that, the pain points here are pretty straight forward -

  1. Being original is hard. Really, really hard.

  2. It’s time-consuming. Not only to generate the content but also to plan it out and schedule them to go live.

When you consider that tweeting many times a day is best practice, you can see how this becomes a burden.

Current Solutions

  • Sushant Kumar's Project

    • Sushant Kumar created an app that generates tweets based on any given word you feed it. He talks openly about the project here. It appears that he’s packaged the tweets up and you can download them all by giving him a monetary donation. I must admit, the Tweets that it produces are surprisingly good.

  • That Can.Be

    • That Can.Be is a tool that uses your existing tweets to generate new tweets. Ultimately, the tool is pretty bad and often creates nonsense tweets. As far as I know, they aren’t using GPT-3 (which is probably why it doesn't work).

  • What to Tweet

    • What to Tweet is a tool that gives you prompts to help you come up with your next tweet. While it does help with writer's block, the tool doesn’t auto-generate tweets. My friend, Jakob Greenfield, launched What to Tweet about a month ago on Product Hunt and it did very well.

Note: You should also look at Copy.ai. They are a great example of a startup built using GPT-3 technology. They don’t offer services for generating tweets, but they could definitely add that to their product in the near future.

How to Execute

Of all the ideas I’ve talked about, this is one that seems relatively easy to pull off:

  1. Find your Early Adopters: To start, focusing on creators/makers that are using Twitter to build an audience. Look for Twitter bios talking about audience building, newsletters, and community builders. Look for people who have paid for guides/courses on growing their Twitter audience. The goal here is to find out if they will pay for this solution.

  2. Create an MVP: As I’ve said in the past, I’m not a technical person. But, it appears building an MVP should be relatively easy. If building on top of GPT-3 is similar to building on top of GPT-2, then there are a number of guides that walk you through building a tweet generator with the technology.

  3. Partner with Twitter Influencers: I’d focus on getting big names to back this early on. Twitter’s platform is built for virality. Getting a few big names in the maker/creator space to endorse this product could help it spread like wildfire.

  4. Look at Business Twitter Accounts: For anyone who has ever managed a social media account for a business, they’ll understand the struggle around content ideation very well. Targeting social media managers could open up access to big budgets and interesting pricing models.

Challenges

  • Getting GPT-3 Access: This is the elephant in the room. Right now, GPT-3 is in closed beta - meaning it’s not open to the public. To gain access, you need to know someone with access or submit an application. That said, it sounds like it’s possible to get access with a bit of hustle and consistent outreach. Alternatively, you could look at building the tool on top of GPT-Neo, an open-source version of GPT-3.

  • Will People Pay for This? The biggest risk I see building this venture is the market risk - will people actually pay for a tool that generates Tweets for them? Luckily, this is easy to find out. You can test this without getting access to GPT-3 or building anything.

  • Low Defensibility: Right now, the only barrier to entry is getting access to GPT-3’s API. Beyond that, there’s nothing stopping competitors from rolling in.

  • Lack of Authenticity: Authenticity is one of the best ways to build an audience. Having AI generate your content could result in backlash from your audience. If content creators are using a tool like this, they may want to remain anonymous.

TL;DR

  • The Idea: A Twitter tool that leverages GPT-3 technology to help you come up with amazing tweets based on your audience and who you follow.

  • The Pain Points: The best way to build an audience on Twitter is to come up with high-quality Tweets at a high frequency, but this is time-consuming and difficult.

  • Opportunity Size: Small ($1M-10M)

  • Why I Love This Idea:

    • It solves a real problem that Twitter users face daily - “tweet generator” gets about 3,000 monthly queries.

    • GPT-3 technology is a perfect solution to the problem.

    • There is little to no direct competition.

  • Biggest Hurdle? Getting access to the GPT-3 API.

Thanks for reading - now get out there and build it! 

Catch ya next week, 

Keevin ✌