Introducing Anansi ❤️🕸
I want you to meet what I've been working on 😄
Hi, Newsletter fam!
It’s been a little while. I’ve been heads down trying to build. To be honest, I was on the fence about writing this post since I’d set a milestone for myself of my following newsletter being the one where I could share access with y’all… but I’m not there yet. Instead, I felt it was more important to share progress in the spirit of building in the open. Just because I’m not production ready doesn’t mean I can’t share.
TL;DR
I’d like to introduce you to Anansi, the interactive storytelling platform I’ve been working on. 😄🕸
Ask: Review the below marketing copy or visit https://www.storiesbyanansi.com and send me your questions or reasons you are interested / not interested.
I’m hoping my next update will be to announce access, but I’m holding off right now because I’m taking advantage of having a developer engaged and polishing the experience.
Without further ado, here’s Anansi (as explained via some draft marketing materials):
Welcome to Anansi, the interactive storytelling platform where your child takes center stage! Our AI-powered platform creates a never-ending world of adventures, encouraging your child to embark on a journey filled with diverse characters and experiences, including their very own friends and family. Each choice leads to a unique outcome, building curiosity and confidence. Anansi's parent-curated content, enhanced by beautiful illustrations, offers a safe and captivating experience to learn.
Step into the magical world of Anansi, where every story choice sparks curiosity and nurtures a lifelong love for reading through fun, interactive adventures.
And here’s a little in-progress video demo:
Why I’m excited
✨ Stories are magical. And I think there’s something super special about putting a child into a story and letting them choose what happens.
🆕 Something like this has never been possible before. Choose-your-own-adventure stories are a classic genre, but they’re time-consuming to build and certainly not personalized.
🚜 I’ve seen some early signs of traction. I’ve had several 20+ min sessions with my 4-year-old daughter using it, and some very early testers also reported long, engaging sessions with their children (despite all the bugs!). My daughter even asked to use it when she had a nightmare to calm down. And 3 of 8 user interviewees proactively and enthusiastically asked me for access with a verbal product description.
🎓It could have some very cool educational value. IMHO, the base product is cool, but it’ll get even cooler if I can weave in some educational material. I have some prototypes of this already working.
What’s been hard
Being patient for the right product. I decided to go all-in and hire a Full-Stack developer via Toptal, and I found Sarah1. Initially, it was to add a few hours of front-end polish to my existing prototype, but as she totally crushed my starter projects, I thought: If I'm going all in, I should have her implement Stripe for payments. In parallel, the feedback I got from some early testers was that the product seemed cool, but the latency was super painful2. So I've been spending the last few weeks improving latency while Sarah's been working on implementing Stripe… and since she's deep in the code base, this is the cheapest it's ever going to be for me to improve the front-end experience to support the near-term features that address early tester feedback. So at the risk of scope creep, I'm delaying availability because of ROI.
Product positioning for real user needs - I was testing for the first 1.5 months to see if I cared enough about Anansi to keep it going. I didn't treat it like a product but as a hobby. I had a few people tell me that I was building a cool tech demo, but I needed to know what problem I was solving, for whom, and testing to see if they would pay. For me, it felt obvious: I hate the children's edutainment products that feel like I'm giving crack cocaine to my child, and I feel like reading stories together is like a delicious healthy meal. But just because I think it's cool doesn't mean it's good for real users. With a mentor's gentle prodding, I finally got around to user research. I found that parents care a lot about ensuring their children's brains are engaged, meeting educational goals, and building curiosity and confidence3. Also, many people don't get it when I explain it. They ask me questions like "How many stories do you need?" or "When does the story end?" not realizing that each story is generated dynamically and a user can generate as many as they want. I'm still struggling to find the proper positioning. Feedback is welcome on my draft marketing pitch. I'm also not 100% sure if the best market is kids 3-7 (with parent facilitation and/or text-to-speech) or kids 8-12 who are reading autonomously. I'm hoping early testing will help me answer that.
I worry about competition. There are a LOT of folks playing in the AI+Children’s stories space. Many of them amount to: Give a user either a) a free prompt text field or b) some pre-canned options, then generate a 1-4 minute long story with some AI-generated images. As far as I know, there are also many choose-your-own-adventure LLM demos, but few have tried to productionize. I keep hoping I’m in the right space and will create something unique at the intersection of my interests and skills, but it worries me.
Things that have surprised me
GPT-4 as a coding buddy and on-hand consultant - I haven’t been to Stack Overflow in weeks. I find it helpful just to put in large chunks of code, and GPT-4 will help me find bugs. I ask it for help with weird build issues. It’s not perfect, and I often call BS, but GPT-4 + GitHub copilot makes my work possible.
AI images are not ready to be unmoderated - The first thing Silicon Valley folks ask me as soon as I explain the concept of Anansi is: “So then you can just hook it up to an AI image generator and generate dynamic images, right?” This siren song has tempted me, but I strongly feel like AI images aren’t ready for this application yet. I’m leveraging Midjourney, and V5/5.1 has been doing a great job of more realistic images (i.e., less body horror), but I feel like zero body horror is the right amount for a kid’s product, and it’s not there yet. Additionally, there’s still a lot of bias. I did a query for “friendly alien” + some stylistic modifiers, and the tool came back with three images of a friendly alien and 1 of a black girl. I was horrified. If Midjourney is generally recognized as the best image generator, AI image generation is still not ready for unmoderated on-demand image generation for kids.
Amplify is not as robust to more complex scenarios - As Sarah and I continue to build out the prototype I built on Amplify, we keep hitting up against Amplify’s quirks and limitations. So I still give Amplify my full endorsement for prototype creation, but I’m lukewarm on a production-level recommendation.
What’s Next
I’ll burn down a final features list and polish them before opening access. 🤞
Based on your feedback and questions, I’d love to iterate on my go-to-market strategy.
May 10 ‘23 Status Update
Success Metrics
After spending Q1 focusing on play, I’m transitioning to holding myself accountable for one metric: Paid users. My Q2 goal will be to get 100 paid users. I have no idea how realistic it is.
Info Diet
Over the last month, there’s been so much great content re: AI. I’m categorizing it into a couple of categories:
General AI innovations or technical explanations
AI art resources
AI for Education
AI Industry News
It's been fantastic working with Sarah so far! I can't recommend her enough. She's got an excellent eye for detail, is willing to have an opinion, offers non-judgemental feedback on my awful code, and delivers quickly with proactive handling of end-to-end use cases. After this engagement, I'm hoping I can work with her again.
Two good resources on experiencing waiting: the psychology & progress bars
Transparency and controls are also a must! But those are table stakes, not a differentiator. But it's good data to know how important these are.