In November 2023, OpenAI launched GPTs which enabling Plus users to customize their own GPTs.
We launched AnotherMind.ai, a platform that empowers non-tech individuals to easily build, chat with, and explore customized chatbots, with curated knowledge and skills, based on their unique needs and interests.
The goal is to explore an intuitive, effortless, and universal operating system that enables everyone to customize and engage with human-curated specialized AIs.
Jennifer is a Product Manager who wants to turn her side project into a startup. She needs to wear multiple hats and venture into unknown domain. Here are 4 bots she built or find to boost her productivity, they are created by herdelf or domain experts for specialized expertise. And there are also bots in PM, Marketing, and Design joining her one-man band.
Following our soft launch and organic outreach, we swiftly accumulated thousands active users in the first week and welcomed our first group of paying customers. However, the feedback also pinpointed a gap; our specialized AI didn’t yet offer a significant efficiency leap (5x or 10x) in many domains, hinting at a need for killer apps. Despite this, we succeeded in empowering non-tech users to craft their AI solutions and explore a range of AI apps.
Moreover, our acceptance into Conviction's Embed Accelerator through AnotherMind.ai brought financial support, GPU resources, and enriching engagements with industry stalwarts. This bolstered our exploratory endeavors, accelerating our pace towards refining and expanding our platform's capabilities, thereby drawing us a step closer to a more user-centric, innovative AI-native operating system.
In this AI summer, I joined AIdeate's founding team, as the first Product Designer, forming a dynamic trio alongside the CEO and CTO. Together, we ventured into the frontier of generative AI technologies.
I built two innovative AI products from idea to MVP: AnotherMind.ai, an LLM platform for customized chatbots, and AIffie, an image generation app for multimedia storytelling.
My responsibilities encompass design tasks including user research, product design, brand design, user testing, and facilitating workshops. Additionally, I wore multiple hats and shared duties in prompt engineering, product, marketing, and community management, etc.
Existing Al platforms often provide over-generic responses, lacking the personalized touch users need for their unique inquiries and tasks. And customizing Al solutions to meet individual needs requires technical expertise. Though the barriar becomes lower, it is still significant for people without domain knowledge in AI and a technical team.
Before Beta launch, we ran an 1-week AI hackathon with 84 target users building on our platform for hypothesis validation.
And we grow user community to keep deep connection with our users.
As we ventured towards conceptualizing an AI-native operating system, a pivotal step was to explore and define the primitives of specialized AI - the fundamental 'Lego pieces' that would constitute the chatbots on our platform. The aim was to modularize these elements in a digestible format for individuals lacking a technical background, posing a significant challenge to innovative design.
Through research, I uncovered 3 core user needs for bots - Consistency, Context, and Capabilities. I explored metaphors to simplify complexity of the technical solutions and had the privilege of garnering feedback from Pinterest co-founder and designer, Evan Sharp, who resonated with simplifying complex technology with metaphors that everyone can understand immediately.
I explored and iterated on the information architecture for the platform through rapid testing. Pinned bots and recent chats were available in the navbar for quick access. Collapsible thread history on the left provided context, while a collapsible details panel on the right, introducing available skills, knowledge, and current instructions - like a profule sidebar. Therefore, I created a scalable pattern for interacting with various specialized AI bots.
A significant issue was users don't resonate with other people's long-tail bots but found them irrelevant. And a bot's profile usually fails to intuitively represent its capabilities.
We shifted focus from just improving bots to matching them with users' specific needs, enhancing discovery. Despite our efforts to categorize bots and improve onboarding, users struggled to find relevant bots. We solved this by revamping search: users input questions, and the system matches bots based on their prompts and knowledge, presenting multiple, relevant answers from different chatbots. Users then choose the best bot for a deeper, personalized interaction.
500+ chatbots created in the first week
First group of paying users
Enrolled into Conviction’s Embed accelerator
Slack integration for teams
We found chatbots’ potentials in turning async communication to sync for teams, streamlining documentation and collaboration.
Multimedia experience & dynamic UIUX
One popular chatbot was an subculture RPG I created with image generation, world view settings, and interactive plots, paving the way for the second product focused on multimedia experience with consistency in character and stories.
The first version of builder is the modularization of Chain of Thought prompt engineering. There are system prompt (customized instruction later by openAI), user input, AI instructions with selected tools and actions like web browsing. And we prototyped it really fast to hand it to our waitlist. The purpose was to find out what users want to create and how to help them achieve it.
1. Prompt Engineering is still hard for the general public as iterating is time consuming. Skills also needs to be auto-generated from conversational interactions. (Explored and requires GPT4 or advanced version later)
2. The knowledge - human-curated data set - is actually the soul of ChatBots. People are looking for unique perspectives. (Priority 1)
2. Creators need Multimedia experience, generative UI/UX, and API to design immersive AI experiences within context. (Priority 2, and explored further in the second product)
