Cation
Your AI companion, minus the AI feeling.
For the past week I've been working on rebranding Knot AI, my AI companion startup.
New logo, new name, new feel. Improved UI, system prompt, fixed bugs, and updated privacy policy and TOS.
I finally finished it today and looking forward for feedback.
A new Feature! You as a user are now able to send multiple messages at the same time and get a accurate response, just like in real chatting apps. The Bond dynamically detects that you are typing another message and silently waits until you have sent another to analyze both in a batch.
llama-3.3-70b-versatile from Groq has been serving well for the development stage but in my experience i think openai/gpt-oss-120b is a better model for this type of product.
Developed a memory evaluation system that runs every once a while and collects important information from your chats and saves it to your Bond's Memory for access later.
Hi IH!
Just launched my first startup as a teen and wanted to share how I got from an idea to a visible product.
The idea
I'd been looking at existing AI companion apps and kept noticing that conversations felt like a one-time experience. Nothing carried forward, everything felt artificial, nothing felt like an actual connection. So I thought why not make Knot focused around making the conversations as human as possible, with features including a mood system, proactive texting (Your Knot messages you first when your away), double/triple texting, and natural language.
Crafting the product
I started working on Knot and after around 2 months of development I finally had a MVP to show to people. Around this development period I had a few burnouts but what I always told myself is that even if this fails there's no pain in building it and gaining some experience rather than searching for another idea only to be paranoid again.
I have to admit that this period was very grindy but if that's what it takes then why not right?
Stack, if you're curious: Next.js + Supabase + Prisma, running on Groq (Llama 3.3 70B) for generation but I do plan on investing in a better paid model once I gather enough audience. All on free-tier infra since I wanted to validate demand before going all in.
Decisions
I did also have to make some crucial decisions while creating Knot. Some were easy and others were difficult to say the least. Looking at my competitors, I knew that they all offered romance/roleplay in some kind of way, be it unfiltered or filtered. So I had to make a decision. I choose to not include any romantic themes or roleplaying just because I think it ruins the conversation and makes my app part of thousands of other that do just the same thing. The quality matters most to me, I wanted to make this app the best I could.
Where it's at now
live, free to use, no subscriptions yet, I'd rather find out if there's real demand before building out a paid tier. Would genuinely love feedback from this crowd, does Knot read as a real differentiator, or does everyone say that about their companion app?
https://knot-companion.vercel.app
Comment