Hacker News Reader

shoucheon Ask HN: How are you building a company-wide MCP strategy?

I would love to hear about the friction points and what your "MCP Roadmap" looks like.

shoucheon Ask HN: Do companies avoid hiring programmers who would rather do indie dev?

Yes, some companies avoid hiring programmers who prefer indie work. They worry these programmers might not fully commit to their job or could develop competing products. Others fear they might prioritize their indie projects over company work.

shoucheon Nvidia reveals Blackwell B200 GPU, the 'most powerful chip' for AI

Explained by AI

https://xplained.vercel.app/ask?q=Nvidia%20Blackwell

Additional details on vRAM and spec: https://xplained.vercel.app/ask?q=Nvidia%20Blackwell%20vRAM%...

shoucheon India test-fires Agni-V ballistic missile with multiple warhead technology

AI explains why Agni missiles are important to India?

https://xplained.vercel.app/ask?q=Why%20Agni%20missiles%20ar...

shoucheon Andrej Karpathy Departs OpenAI

https://xplained.vercel.app/ask?q=Karpathy%20left%20OpenAI

AI explains the update

shoucheon Ask HN: What have you built with LLMs?

Created an AI explainer app that helps you understand a topic, kind of like Perplexity.

It's currently free to use. Its built using nextjs+tailwind and is powered by Vercel + Brave + Gemini Pro. https://xplained.vercel.app

There are other projects that I worked on as part of my job, mostly around bots, search, classification, and analytics.

shoucheon Ask HN: Seeking guidance on selling my hobby project on Acquire.com

Hey HN community,

A few weeks ago, I created Xplained, an AI-based answering engine that uses web data and LLM to provide responses (like perplexity). I am planning to sell my hobby project on Acquire.com and could use your advice.

What are your experiences with selling projects? How did you determine the valuation, pitch to buyers, and ensure a smooth sale process?

I'm looking for practical tips and insights to navigate this transition effectively.

Appreciate your guidance!

shoucheon Show HN: Write Documentation Easily

Looks like a Novel.sh wrapper. It would be useful if it could have templates by use cases like product document or customer support or FAQ or product listing. Also, maybe some AI assistant to help with content writing.

shoucheon Show HN: EnhancedJournal – Elevate Journaling with AI Insights

Hey, any sample that I can see? As a user I feel apprehensive to sign-in before knowing what I would get.

Also, UI can be improved for mobile.

shoucheon Show HN: Knowledge Sharing Protocol via semantic lookup with use in LLMs

As you mentioned, if LLMs are not required, I don't see the novelty - it basically is search on embeddings.

GPT Assistants readily offer this, with options to customize.

shoucheon Show HN: An interactive visualization of a semantic cache

Hey. Its a good practical application, especially to reduce cost. But at about 0.6 similarity, I get some cache hit. Maybe with more examples and for a high use app, the cache hit would increase based on a higher similarity scores.

Still early days on this I guess, but any observations that can help improve the hit rates?

shoucheon Show HN: Attabit – AI powered news

I liked the idea. A good start to a promising app. It does need some features to make it useful and some other improvement ts on UX. People have listed others, but I would like a filter or search at least for starters. Also, progressive disclosure on the news details.

On the cost side, you could try Gemini Pro, currently free with usage limitations. But since the content is saved, it should be fine.

How are you sourcing news and deciding what news to keep? Search engine or some kind of feedback? Would like more details here.

shoucheon Show HN: Xplained – AI-Driven Instant Explainer (Perplexity inspired)

Updated and significantly improved the app.

shoucheon Show HN: Xplained – AI-Driven Instant Explainer (Perplexity inspired)

Hi. Thanks for the feedback.

Its just a quick POC made in a day. I am still working on integrating a backend and optimizing the queries. Also, it currently uses free APIs with small usage limits. Plan to upgrade soon.

shoucheon Ask HN: What are you using to iterate and test your prompts for LLMs?

OpenAI came up with its guidelines on prompting. I created a GPT assistant to help me improve the prompt based on that. https://chat.openai.com/g/g-haH111AXX-prompt-optimizer

The results are good, and improve the output quality significantly. Small manual QC and tweaks further improve the responses.

shoucheon Ask HN: How do you find ChatGPT for more complex coding

Iterations and assigning roles to the GPT help. Give a small block of code, ask GPT to review this with a panel of developers (frontend, backend, ...), users and UI/UX experts for 5 iterations, and then provide the final code. GPT 4 works better. However, a lot of times the code is outdated and creates new issues. Manual review and fixes are needed for most custom code blocks. It works well for standard code modules.

shoucheon Making Music from Images in Python

Hey. I had the same idea as you a while back. Did some samples as well. Used a pretty similar approach till the "Generating Other Scales" section. But was not happy with the output. However, I found them as a good seed to generate music using AI systems from that time. The current systems should provide better results.

Here are some examples: https://opensea.io/assets/matic/0x2953399124f0cbb46d2cbacd8a...

https://opensea.io/assets/matic/0x2953399124f0cbb46d2cbacd8a...

https://opensea.io/assets/ethereum/0x495f947276749ce646f68ac...

shoucheon Build a search engine, not a vector DB

I have been using elastic index for a while now. The best way I have found is to use a hybrid search - match all with embedding + exact+fuzzy match combination as a way to boost results.

Reranking also provide a significant improvement to the response quality.

Another way to improve results for domain specific RAG systems is to use some heuristics to boost results. E.g., penalize results that contain certain negative keywords or boost results with certain patterns.

For RAG, given the limited context size and potential hallucinations, best prompt + best data will provide you with best response.

Prompts can be improved greatly to get the LLM to throw a good response with reduced hallucinations. A lot of techniques are seen on Twitter and can be explored to find a good fit.

I improve my prompts using a GPT assistant that significantly improve the response quality. https://chat.openai.com/g/g-haH111AXX-prompt-optimizer

shoucheon Ask HN: Is Twitter/X working for you?

Just showed follow suggestions, everything else is down