HNHacker News
TopNewBestAskShowJobs

zh2408

470 karma · joined December 8, 2023

submissionscomments
zh2408··on LLM Agents Are Simply Graph – Tutorial for Dummies
Thank you!!
zh2408··on LLM Agents Are Simply Graph – Tutorial for Dummies
Hey folks! I just posted a quick tutorial explaining how LLM agents (like OpenAI Agents, Pydantic AI, Manus AI, AutoGPT or PerplexityAI) are basically small graphs with loops and branches. For example:

OpenAI Agents: for the workflow logic: https://github.com/openai/openai-agents-python/blob/48ff99bb...

Pydantic Agents: organizes steps in a graph: https://github.com/pydantic/pydantic-ai/blob/4c0f384a0626299...

Langchain: demonstrates the loop structure: https://github.com/langchain-ai/langchain/blob/4d1d726e61ed5...

If all the hype has been confusing, this guide shows how they actually work under the hood, with simple examples. Check it out!

https://zacharyhuang.substack.com/p/llm-agent-internal-as-a-...

zh2408··on Show HN: I built Cursor Agents with Cursor (Open Source)
I just released a simple, open-source version of Cursor Coding Agents. You give it a user request and a code base, and it’ll explore directories, search files, read them, edit them, or even delete them—all on its own!

I built this using the leaked Cursor system prompt (plus my own guesses about how Cursor works). It’s missing a few features like code indexing, but it still runs smoothly on the latest Sonnet 3.7 thinking model. Everything is minimal and fully open sourced, so you can customize it and add your own knowledge. If you’re curious about how I built it, I made a full step-by-step video tutorial. Enjoy!

zh2408··on Show HN: Ask AI Paul Graham (Open Sourced)
HAHAHA thank you!
zh2408··on Show HN: I built an agent to write personalized cold email opener
For cold email, the first few sentences are make-or-break - they determine whether someone keeps reading or hits delete. I built a small LLM app that:

1. Takes a person’s name + a few keywords 2. Does a Google search to find relevant info about them 3. Identifies personalization opportunities, such as: Personal connections (same alma mater) Recent achievements (promotions, awards) Content they've created (talks, articles) Hobbies and interests 4. Crafts a personalized opener based on your preferred tone (friendly, professional, etc.)

You can check out the repo here: https://github.com/cold-outreach/Tutorial-Cold-Email-Persona... And here’s a live demo if you want to see it in action: https://pocket-opener-851564657364.us-east1.run.app/

This is built with Pocket Flow, a 100-line LLM framework: https://github.com/The-Pocket/PocketFlow

zh2408··on Show HN: Demo of my web game about social persuasion
Very cool!
zh2408··on YC is wrong about LLMs for chip design
Human itself is like a slow, expensive, non-deterministic black box...
zh2408··on Show HN: RL Agent that can auto-optimize your LLM prompts
1. How does it compare to DSPy? 2. Some enterprise tasks have difficulty finding good metrics? Do RL fails in these cases?
zh2408··on Launch HN: Codebuff (YC F24) – CLI tool that writes code for you
How does it work if I'm not adding features, but want to refactor my code bases? E.g., the OOD is poor, and I want to totally change it and split the codes into new files? Would it work properly as it requires extensive reads + create new files + writes ...
zh2408··on Launch HN: Codebuff (YC F24) – CLI tool that writes code for you
This is so REAL: LLMs suck probably because your modularity sucks LOL
zh2408··on Launch HN: Midship (YC S24) – Turn PDFs, docs, and images into usable data
Saw reducto released benchmark related to your product: https://reducto.ai/blog/rd-tablebench Curious your take on the benchmark and how well midship performs
zh2408··on Show HN: GoSQL – A query engine in 319 LoC
Feel like you can achieve something similar in duckdb? duckdb allows you to query local csv, parquet, and even remote ones?
zh2408··on Show HN: Compiling C in the browser using WebAssembly
What's the use case?
zh2408··on Show HN: An open-source tool that cleans tables using LLMs
The project is open sourced at: https://github.com/Cocoon-Data-Transformation/cocoon
zh2408··on Show HN: An open-source tool that semantically profiles your data using LLMs
Thank you so much! I'll fix the warning. Yes, that's an issue with LLMs. I'm using Claude 3, which was trained on data up to August 2023. I'll add the current date in the prompt. Thanks!
← PreviousPage 2 of 2