Listen to this story
An India-based developer, who goes by the pseudonym Kamban, has developed an AI writing app for Mac called Elephas. It uses the GPT-3 model to help users write all kinds of content, including business documents, long-form articles, emails, social media posts, and even product reviews. Analytics India Magazine got in touch with Kamban to learn more about this tool.
Elephas – AI writing tool
There are many copywriting tools like Copy AI and Content AI, but most of them work on the premium model and do not offer the ‘power of GPT-3’, as Kamban puts it. “I built a native app for Mac called Elephas using GPT-3. It works on top of all your Mac applications, and no need to switch windows. I wrote the code in Swift programming language. This app takes your OpenAI keys and uses them for all communications. This is one of the major differences when you compare it with the other AI writing tools, where they usually store your information. Elephas does not send your text information to our servers,” Kamban said.
An AI-based writing cum research tool for business professionals, the way Elephas works is that the user needs to select a text and options from the menu bar (for example – rewrite). Elephas accesses the selected text and converts it into the requested format. Kamban initially just built a simple wrapper and posted a sample of it on the discussion forum HackerNews. “It kind of exploded, and Elephas was even trending on the homepage of HackerNews for some time,” he said. He admits that even though back then the version did not have a great look, it helped him gain valuable feedback and a few paid customers.
Sign up for your weekly dose of what's up in emerging technology.
Short product demo (2 mins) : https://vimeo.com/771041299
The next step was taking end-user feedback. “What I introduced initially was just a wrapper. Now, there is a utils section in the app, which you can use to generate from Google Sheets formulas to presentations. Let’s say you want to write a presentation on the impact of air pollution in India – just give the title, and it will give you the ten presentation slides. Users can finetune the presentation to best suit their objectives,” explained Kamban.
Download our Mobile App
Kamban adopted several methods to optimise the app. “I reviewed the major features our initial users used and adjusted the landing page to reflect them, which resulted in better conversions. I realised a lot of my users did not understand the full potential of the tool. I set up a knowledge base tool and organised all the features in an easy-to-navigate structure. I also created product demo videos, completely created by AI. With these, the support requests dropped,” he said.
Kamban has been building side projects for four years and has never made any revenue. For the first time, the Elephas app has reached $1000 MRR recently over a period of five months. Elephas has over 200 business professionals and content writers using it every day.
Playing with AI
Elephas is not the only product that Kamban has built.
One of his projects called SwanSearch – a search engine built for developers to provide them with inplace answers – features code search, function search, and JSON validation apart from viewing answers on the same page without needing to visit other websites in most cases. FlatGA is another active project that Kamban is maintaining, which unifies all website metrics like analytics, SEO, monitoring, and performance into a single tool.
His other projects include browser tab manager, an AWS cloud savings tool, and a GPT-3 based well-being assistant. Most of his tools are based on GPT-3. “GPT-3 offers features that I couldn’t get anywhere else. For example, it can produce remarkable output for new inputs, which in the past would have needed a huge amount of training data and computing power. GPT-3 reduced the entry barrier for a lot of my projects,” he said.
He further added, “If you notice, businesses are already hiring for a job called ‘Prompt engineers’. We are only getting started with advanced LLMs. GPT-4 could fix some of the shortcomings of GPT-3, such as maintaining factuality, increased memory retention, and making good external references.”
Kamban places a lot of trust in the advancement of AI, especially large language models. He said that LLMs are evolving faster than any other AI, which will have a huge impact on the way the world operates – such as more personalised and adaptive learning experiences and increased use of data and analytics to drive decision-making.
He concluded that his ultimate goal is making AI accessible to regular people. “I will focus on providing platforms where people can easily use AI’s potential,” he signed off.