Show HN: I'm 13 and built an AI that remembers context across conversations(ai.nityasha.com)
ai.nityasha.com
Show HN: I'm 13 and built an AI that remembers context across conversations
https://ai.nityasha.com
17 comments
This is pretty cool!
I realize that you (the 13 year old) probably did not walk this path alone and I'm guessing you had a lot of help. Not sure how much help, but I don't mind giving you the benefit of the doubt today.
Great to see that you're being creative and (at least) participating in making things!
I know that it took a lot of work for my parents to encourage my passions and for me to actually pursue them in earnest. I could have easily gotten lost in being a consumer of media and programs. Instead, I make them and it is very fulfilling. I now have a lot to thank them for and I don't think I'd be here without all their hard work and hoisting me up in my youth. It sounds like you have parents that care that much and for that, you should be grateful and participate in your growth as they do.
I realize that you (the 13 year old) probably did not walk this path alone and I'm guessing you had a lot of help. Not sure how much help, but I don't mind giving you the benefit of the doubt today.
Great to see that you're being creative and (at least) participating in making things!
I know that it took a lot of work for my parents to encourage my passions and for me to actually pursue them in earnest. I could have easily gotten lost in being a consumer of media and programs. Instead, I make them and it is very fulfilling. I now have a lot to thank them for and I don't think I'd be here without all their hard work and hoisting me up in my youth. It sounds like you have parents that care that much and for that, you should be grateful and participate in your growth as they do.
Thank you for the kind words! You're absolutely right - I didn't do this alone. My dad has been incredibly supportive throughout this journey, helping me learn to code and working alongside me on this project. The technical architecture and many of the hard problems were definitely team efforts.
I'm grateful to have that support system. It's easy to just consume content and games, so having someone push me to create instead has made all the difference. Your parents sound like they did the same for you - that's awesome.
Thanks for taking the time to share your experience!
I'm grateful to have that support system. It's easy to just consume content and games, so having someone push me to create instead has made all the difference. Your parents sound like they did the same for you - that's awesome.
Thanks for taking the time to share your experience!
I don't think you should announce your age on a public forum. Generally, no one should at any age (I break this rule, but no one is trying to groom middle-aged men). At your age, definitely don't.
Kill your account and start fresh with better opsec. You're inviting predators.
Oh, I only just scanned the description and it sounds like you're likely female (to a westerner, anyway). Holy crap, kill this account and never link back to it for the next decade or more.
Kill your account and start fresh with better opsec. You're inviting predators.
Oh, I only just scanned the description and it sounds like you're likely female (to a westerner, anyway). Holy crap, kill this account and never link back to it for the next decade or more.
You're right to call this out - I appreciate the concern. I didn't think through the safety implications of sharing personal information publicly.
I'll be more careful about operational security going forward. Thank you for the direct feedback.
I'll be more careful about operational security going forward. Thank you for the direct feedback.
[flagged]
Fair point! I could have done a better job explaining the technical depth upfront rather than leading with my age.
What we built: persistent memory using vector embeddings (Pinecone), semantic search across conversations, Socratic teaching system, and unified workflow integration. The value isn't in the base model - it's in the architecture layer we added on top.
I should've let the technical work speak first. Appreciate the feedback.
What we built: persistent memory using vector embeddings (Pinecone), semantic search across conversations, Socratic teaching system, and unified workflow integration. The value isn't in the base model - it's in the architecture layer we added on top.
I should've let the technical work speak first. Appreciate the feedback.
[stub]
So a wrapper around an existing LLM? even chatgpt can reference previous conversations and remember context.
Also the fake comments (or friends comments?) are not really liked on Show HNs.
Congrats on building it.
Also the fake comments (or friends comments?) are not really liked on Show HNs.
Congrats on building it.
Fair point on the "wrapper" label. Let me clarify what we're building on top of base LLMs:
Yes, we use OpenAI/Anthropic APIs - we're not training models from scratch (like you said, neither does Perplexity, Jasper, or most AI tools).
What we add (technical details):
1. Persistent Memory Architecture - Vector embeddings of user context stored in Pinecone - Semantic search across past conversations (not just in-session) - Retrieval pipeline: query → embed → cosine similarity → top-k memories → inject in prompt - Challenge: Managing token costs while maintaining context
2. Socratic Teaching System (Study Mode) - Question analysis: detect knowledge gaps - Progressive hint generation (not just Q&A) - Tracks learning progression - Example: Instead of "here's binary search code", asks "what property of sorted arrays makes this possible?"
3. Unified Workflow Integration - Email parsing + calendar sync + task extraction - Single interface reduces context switching - Memory persists across all tools
Architecture overhead: - User sends query - Retrieve relevant memories (vector search) - Build augmented context window - Send to LLM with enriched prompt - Generate + store new embeddings - ~200ms additional latency for memory operations
You're right that the base intelligence is GPT-4/Claude. But saying "wrapper" feels like saying Notion is "just a wrapper around PostgreSQL" or Stripe is "just a wrapper around payment processors."
The value is in the layer we built, not the underlying model.
That said - we should've been clearer about this upfront. Our first comment didn't explain the technical depth. That's on us.
Re: the fake comments - you're absolutely right to call that out. Those were friends/early users we asked to support the launch. That was a mistake and goes against HN's culture of authentic discussion.
I'm 13 and this is our first HN launch. We didn't understand how much the community values genuine engagement over orchestrated support. Won't happen again.
Apologies to the HN community for trying to game the system. Should've let the product speak for itself.
Appreciate you taking time to give honest feedback instead of just downvoting. This is exactly why we launched here - to learn from people who know better.
What would make this feel genuinely useful vs "just another wrapper" to you?
Yes, we use OpenAI/Anthropic APIs - we're not training models from scratch (like you said, neither does Perplexity, Jasper, or most AI tools).
What we add (technical details):
1. Persistent Memory Architecture - Vector embeddings of user context stored in Pinecone - Semantic search across past conversations (not just in-session) - Retrieval pipeline: query → embed → cosine similarity → top-k memories → inject in prompt - Challenge: Managing token costs while maintaining context
2. Socratic Teaching System (Study Mode) - Question analysis: detect knowledge gaps - Progressive hint generation (not just Q&A) - Tracks learning progression - Example: Instead of "here's binary search code", asks "what property of sorted arrays makes this possible?"
3. Unified Workflow Integration - Email parsing + calendar sync + task extraction - Single interface reduces context switching - Memory persists across all tools
Architecture overhead: - User sends query - Retrieve relevant memories (vector search) - Build augmented context window - Send to LLM with enriched prompt - Generate + store new embeddings - ~200ms additional latency for memory operations
You're right that the base intelligence is GPT-4/Claude. But saying "wrapper" feels like saying Notion is "just a wrapper around PostgreSQL" or Stripe is "just a wrapper around payment processors."
The value is in the layer we built, not the underlying model.
That said - we should've been clearer about this upfront. Our first comment didn't explain the technical depth. That's on us.
Re: the fake comments - you're absolutely right to call that out. Those were friends/early users we asked to support the launch. That was a mistake and goes against HN's culture of authentic discussion.
I'm 13 and this is our first HN launch. We didn't understand how much the community values genuine engagement over orchestrated support. Won't happen again.
Apologies to the HN community for trying to game the system. Should've let the product speak for itself.
Appreciate you taking time to give honest feedback instead of just downvoting. This is exactly why we launched here - to learn from people who know better.
What would make this feel genuinely useful vs "just another wrapper" to you?
We've moved all the comments from new accounts to a stub to prevent them from gumming up the thread.
Abir-01(1)
Ayish_Shaikh(1)
I started coding at 9 on a 4GB RAM laptop. We failed 8 times before this—coupon sites, freelancing platforms, consulting. Nityasha is different: it uses Thesys generative UI for visual charts, includes Study Mode with Socratic teaching, and integrates everything so you don't need 10 tabs open.
500+ active users now. We just launched Nityasha Connect where businesses can integrate services directly into the AI.
Would love your feedback!