Hacker News Reader

csjhon Moonbit: Fast, compact and user friendly language for WebAssembly

Probably smart enough, since code size is part of their mission statement

csjhon Moonbit: Fast, compact and user friendly language for WebAssembly

I’ve been considering trying my hand at a C-like WASM language, with most features mapping directly to WASM instructions. Surprised no one’s done something like that yet. Very cool tech

csjhon Moonbit: Fast, compact and user friendly language for WebAssembly

I think the Rust and Go comparisons are because they’re popular languages with first class WASM support but I agree

csjhon How Is LLaMa.cpp Possible?

IMO the direction we're going seems more like having a few small models in a MoE that are equivalent to a current 200bn model

csjhon Svelte 5 is going to be radical

Isn't that just SolidJS?

csjhon Svelte 5 is going to be radical

Learning Svelte 4 is no different than learning Svelte 3, the changes are almost entirely behind the curtains. Apparently Svelte 5 will also be very minimal breaking changes, so any knowledge will pass over trivially.

In other words, invest all your time in learning svelte :>

csjhon StableCode

phi-1[0] is only 1.3 billion parameters and performs very well in coding tasks - small models have a massive amount of potential

[0] - https://arxiv.org/abs/2306.11644

csjhon Val, a high-level systems programming language

C++ has proper package management?

csjhon PdfGptIndexer: Indexing and searching PDF text data using GPT-2 and FAISS

Why have the OpenAI dependency when there's local embeddings models that would be both faster and more accurate?

csjhon Functions and algorithms implemented purely with TypeScript's type system

To be fair it’s fun to write ridiculous types to get ridiculously specific intellisense

csjhon CSS-Only Chat

rich harris' dream world

csjhon Svelte 4

that's just every developer switching every single one of their side projects to svelte in excitement

csjhon Svelte 4

The breaking changes aren't really gamechanging stuff, just cleaning up baggage in preparation for Svelte 4.x and Svelte 5. The course would still be fully applicable

csjhon Stackoverflow is investing into baking GenAI

Duplicate question closing will now be instant if your question has an embedding that's remotely close to another question

csjhon Gorilla: Large Language Model Connected with APIs

Still outputs natural language. Example from their colab:

Input: I would like to translate from English to Chinese.

Output:

<<<domain>>>: Natural Language Processing Text2Text Generation

<<<api_call>>>: M2M100ForConditionalGeneration.from_pretrained('facebook/m2m100_1.2B')

<<<api_provider>>>: Hugging Face Transformers

<<<explanation>>>: 1. Import M2M100ForConditionalGeneration and M2M100Tokenizer from the transformers library.

2. Load the pre-trained M2M100 model and tokenizer using the from_pretrained() method. The model is trained to translate text from English to Chinese, among other languages.

3. Encode the input text in English using the tokenizer.

4. Generate the translation using the model.generate() method.

5. Decode the output tokens using the tokenizer to obtain the translated text in Chinese.

6. Print the translated text.

csjhon Zap – Fast backends in Zig

It's still pretty early on, but it's aim is to be a C successor I believe. Main example of usage off the top of my head is for bun.sh

csjhon Llama.cpp: Full CUDA GPU Acceleration

I believe it's moreso for the (actively pursued) speed optimizations it provides. When inference is already computationally expensive any bit of performance is a big plus

csjhon Llama.cpp: Full CUDA GPU Acceleration

They come a dime a dozen on HuggingFace, check out https://old.reddit.com/r/LocalLLaMA/wiki/models for a few options

csjhon Apollo will close down on June 30th

They say somewhere that the 20m is from opportunity cost, so 10m for an app that's "costing" them 20m a year would be a deal.

csjhon Texas professor fails entire class from graduating- claiming they used ChatGTP

All of my CS classes already have rules against AI (though they had a head start from copilot). A couple other non-STEM courses as well.

csjhon Htmx Is the Future

SvelteKit forms work fine with Javascript disabled -- Rich Harris is a big proponent of progressive enhancement.

csjhon Launch HN: Pynecone (YC W23) – Web Apps in Pure Python

https://krausest.github.io/js-framework-benchmark/index.html

My personal favorite is Svelte

csjhon Removal of Heroku free product plans

They have crazy low bandwidth limits though... puts me off quite a bit

csjhon Removal of Heroku free product plans

Render has pretty good free tier benefits, but no git push render master

csjhon Issues with upstream DNS provider

probably worth staying at render considering how many issues heroku's had recently