pcwelder

7 ms·

61.
by pcwelder 2y ago
The repo was private, updated. Thanks!!
62.
by pcwelder 2y ago
Gemini 2.5 pro hasn't been as good as Sonnet for me. The prompt I have tried repeatedly is creating a react-vite-todo app. It doesn't figure out tailwind related issues. Real chats: Gemini: https://github.com/rusia
63.
by pcwelder 2y ago
While vibe coding you're code blind, but you can still be aware of the architecture. If all the subsystems are simple and sparesly connected, any bug will likely be isolated and easy to detect. You can build complex systems using simpl
64.
by pcwelder 2y ago
Awesome! I essentially started with the same premise when coding "wcgw" [1]. Give a single shell tool to chatgpt (later claude). Chronology that followed: 1. Some processes take long time: add status checking. 2. Need to kill long
65.
by pcwelder 2y ago
Thanks for sharing this. I got curious and I wondered if a similar thing could be done using pexpect. I went ahead and asked claude to create an MVP. Works pretty well for 20 mins worth of work. Pexpect is a great tool imo. If anyone is cur
66.
by pcwelder 2y ago
We'll have to look with fresh eyes on the print debugging vs debugger debate for LLMs. LLMs can not only add tons of log statements fast but parse the large output from a program faster. How exactly is debugger connection better for LL
67.
by pcwelder 2y ago
Last week I nuked my home directory by running `rm -rf *` mistakenly from the wrong directory. Lost around 2-3 hours to re-install and re-clone. Accidents will happen, disaster recovery is needed notwithstanding misaligned AI. IMO once you&
68.
by pcwelder 2y ago
I connected Chatgpt app to termux to have shell access using custom gpts. The thing is you can talk to custom gpts over voice. There's day and night difference compared to using mobile keyboard to type for writing commands and files. I
69.
by pcwelder 2y ago
Encapsulation isn't the real reason why types help. Pass 5 variables individually, the only difference will be in verbosity. The real benefit is type narrowing. (Or declaring the space of all possible combinations of values.) Instead
70.
by pcwelder 2y ago
- Browser control tools - Shell control tools - File edits tools - Task management workflow/tools - Memory management workflow/tools - Subtasks handoff to other agents Anything major missing raw capability wise?
71.
by pcwelder 2y ago
Claude code terminal ux feels great. It has some well thought out features like restarting conversation with compressed context. Great work guys. However, I did get stuck when I asked it to run `npm create vite@latest todo-app` because it n
72.
by pcwelder 2y ago
Good work. I wonder if there's really a need for separate write to terminal and read output functions? I was hoping that write command itself would execute and return the output of the command, saving back and forth latency. > and i
73.
by pcwelder 2y ago
> A wild thing about LLMs is that they can observe what they've done and decide whether they've done a good job. I've seen this behaviour in Claude but don't remember 4o doing the the same or as frequently.
74.
by pcwelder 2y ago
Most the problems you mentioned will likely be solved with the next iterations of Devin or similar product. I can say that because I work daily with Claude as an agent over mcp, and the problems you mentioned feel very familiar. Based on th
75.
by pcwelder 2y ago
Was the comparison done with or without code context (as obtained using RAG or letting Sonnet ask for files)?
76.
by pcwelder 2y ago
A whole article without mentioning the name of the LLM. It's not like Sonnet and O1 have the same modalities. Anything you do today might become irrelevant tomorrow.
77.
Syntax-checker for multiple languages in a single Python library(github.com)
1 points by pcwelder 2y ago|0 comments
78.
by pcwelder 2y ago
For building mcp server? The official docs do a great job https://modelcontextprotocol.io/introduction My own mcp server could be an inspiration on Mac. It's based on pexpect to enable repl session and has some tricks
79.
by pcwelder 2y ago
Code execution is okay, but soon runs into the problem of missing packages that it can't install. Practically, sandboxing hasn't been super important for me. Running claude with mcp based shell access has been working fine for me,
80.
by pcwelder 2y ago
First one is due to me asking it to return a 4 tuple instead of a list for the rgba_255 specifically, I guess it didn't update Readme or other return values. The second is an artefact of a test case failing, which it tried to fix it us
81.
by pcwelder 2y ago
No everything in the repo is AI generated.
82.
by pcwelder 2y ago
> 3. The "Trust but verify" pattern To add on to this point, there's a huge role of validation tools in the workflow. If AI written rust code compiles and the test cases pass, it's a huge positive signal for me, becau
83.
Computer use on Claude desktop app using model context protocol(github.com)
1 points by pcwelder 2y ago|1 comments
84.
by pcwelder 2y ago
I was curious how well Claude app performs on computer use tasks. So I integrated (screenshot, keyboard, mouse, shell, file tools) with claude using model context prompt. While it's good as a coding and shell agent, the computer use ab
85.
by pcwelder 2y ago
It's great! I quickly reorganised my custom gpt repo to build a shell agent using MCP. https://github.com/rusiaaman/wcgw/blob/main/src/wcgw/client/... Already getting value out of it.
86.
Claude desktop shell agent using Model Context Protocol(github.com)
3 points by pcwelder 2y ago|1 comments
87.
by pcwelder 2y ago
I have written this personal utility to help Claude access my shell execute commands, edit code, run projects, etc. Sharing in case it helps others as well. It's built on the code base that enables ChatGpt web/desktop to talk to t
88.
by pcwelder 2y ago
I just want to thank people behind uv. The tool is just amazing for development, packaging and running packages. And it's blazing fast!
89.
by pcwelder 2y ago
The simplicity of golang definitely helps. Between golang and rust, chatgpt has much easier time with golang. The standard toolchain (as opposed to python packaging) makes it easier for LLM to debug and run commands. So it's easier to
90.
by pcwelder 2y ago
> state of the art in speech recognition for decades Decades doesn't sound right. Around 2019, the Jasper model was SOTA among e2e models but was still slightly behind a non e2e model with an HMM component https://arxiv

More ›