tarruda

18 ms·

271.
by tarruda 3y ago
It is more a matter of trust than effort, eg being less exposed to supply chain attacks.
272.
by tarruda 3y ago
I think the main appeal of projects like Bun and Deno is the built-in tooling for building/bundling modern typescript applications without requiring dozens of dependencies for even a basic hello world app. If node.js decided to include
273.
by tarruda 3y ago
What is the state of Deno's node.js compatibility when compared to Bun?
274.
by tarruda 3y ago
systemd-nspawn is simple but AFAIK it doesn't do any security other than the kernel namespacing. Docker is even worse because it runs containers as root, which means a rogue process can take over the host very easily. Incus/LXD ru
275.
by tarruda 3y ago
For Linux I highly recommend Incus/LXD. Launching a VM is as simple as ``` incus launch images:ubuntu/22.04 --vm my-ubuntu-vm ``` After launching, access a shell with: ``` incus exec my-ubuntu-vm /bin/bash ``` Incus&#x2F
276.
by tarruda 3y ago
> it's not eating crap. Add "and regularly exercising" to the formula. Our bodies are not made to be sitting all day.
277.
by tarruda 3y ago
Nothing special other than llama.cpp, which is an inference engine optimized for apple silicon. I heard you can simply install ollama app which uses llama.cpp under the hoods, but has a more user friendly experience.
278.
by tarruda 3y ago
> You would need multiple GPUs with shared memory if you wanted to offload the higher precision models to VRAM. Or just a powerful apple silicon machine? I've tried dolphin mixtral 4bit on a 36gb ram MacBook m3, and inference is sup
279.
by tarruda 3y ago
It is based on Mistral which llama.cpp supports, so I assume it does run (you might need to convert to GGUF format and quantize it).
280.
by tarruda 3y ago
Hi Andrew. I once saw a video (at least I think it was a video) where you suggested the possibility of Zig reaching language stability in 2025. Is that still accurate?
281.
Chronosphere acquires Calyptia to extend its observability platform(techcrunch.com)
6 points by tarruda 3y ago|0 comments
282.
by tarruda 3y ago
An easy way is to use an OpenAI compatible server, which means you can use any GPT plugin to integrate with your editor
283.
by tarruda 3y ago
> but the larger model is a lot slower. I found the performance to be very acceptable for 33b 4 bit on a m3 max with 36gb ram (much faster than reading speed)
284.
by tarruda 3y ago
Becoming an indie game developer on spare time could make you happy
285.
by tarruda 3y ago
4-bit quantized 33b runs great on a mp pro with m3 max chip
286.
by tarruda 3y ago
AFAIK deepseek coder family are the best open coding models. I haven't tested, but I think deepseek coder 33b can run in a single RTX 3090 when 4-bit quantized. In your case you might be able to run the non quantized version
287.
by tarruda 3y ago
Deepseek coder 6.7B is very useful for coding and can run in consumer GPUs. I use the 6bit GGUF quantized version on a laptop RTX 3070
288.
by tarruda 3y ago
Note that they don't compare with deepseek coder 6.7b, which is vastly superior to much bigger coding models. Surpassing codellama 7b is not that big of a deal today. The most impressive thing about these results is how good the 1.3B d
289.
by tarruda 3y ago
I'm skeptical that there's any JIT for any programming language that can match the raw performance and memory efficiency of LuaJIT
290.
by tarruda 3y ago
> employees take fewer risks Like going to the masseur during working hours?
291.
by tarruda 3y ago
No. Mixtral Medium is one of the proprietary models by Mistral AI, and only accessible via their API: https://docs.mistral.ai/ It currently ranks 4 in chatbot arena leaderboard (slightly behind GTA-4 ELO rating): https:&#x
292.
by tarruda 3y ago
I will keep that in mind, thanks. My initial research suggest that Godot games are written in C#. Since you use C/SDL2, do you code directly your own OpenGL engine?
293.
by tarruda 3y ago
I want to begin game development as a hobby, but I'm unsure where to start. I did follow through https://learnopengl.com/ a few years ago, and while it was a very interesting experience, I imagine I would need to use a
294.
by tarruda 3y ago
Conclusion in the blog post says it all: > I regret to say it, but it's true: most of today's programming consists of regurgitating the same things in slightly different forms. High levels of reasoning are not required. LLMs ar
295.
by tarruda 3y ago
Anecdote: My grandma is 92 (will be 93 in February), has always eaten a lot and never did any exercises except walking on the street. She used to be a school teacher and walked to work. I don't know how many calories she takes, but alw
296.
by tarruda 3y ago
Instead of being a 10x engineer, dial down your performance to match the rest of the team. Then, use the extra time/energy to work on side projects that make you happy.
297.
by tarruda 3y ago
You're right, llama-cpp-python OpenAI compatible endpoint works with `stream:true` and you can interrupt generation anytime by simply closing the connection. I use this in a private fork of Chatbot-UI, and it just works.
298.
by tarruda 3y ago
Also not familiar with llamafiles, but if it uses llama.cpp under the hoods, it can probably make use of mmap to avoid fully loading on each run. If the GPU on Macs can access the mmapped file, then it would be fast.
299.
by tarruda 3y ago
As someone who only used sed for quick pipeline editing, I'm curious of what constitutes a "sed project" that this tool aims to build. Does anyone have examples?
300.
by tarruda 3y ago
In 2023 I had to switch my work web browser to Brave because Slack decided to stop working on Firefox.

More ›