ive developed some low data use habits from using mostly mobile data for internet.

i use newsraft to download the podcasts and videos I listen to, to a VPS using their fast and nearly unlimited internet.

then i use ffmpeg to convert them on the server to 24b opus files before downloading them to my local machine.

for audio files, it generally reduces the file size to 1/3. for video files, the reduction is significantly larger.

then i push the files to an oldschool iPod so i can listen and do stuff at the same time.

works for me.

What I hate is exactly what you mentioned. There's a whole ecosystem of bad frameworks and your job is to find the least bad one. For me it's react/typescript/webpack etc. it took me so long to hack it together. I think I may have actually cried at one point.

I hope nobody here is using leverage... #btc

do you trust any of us ? don't

Llama 4 is out. 👀

Llama 4 Maverick (402B) and Scout (109B) - natively multimodal, multilingual and scaled to 10 MILLION context! BEATS DeepSeek v3🔥

Llama 4 Maverick:

> 17B active parameters, 128 experts, 400B total parameters > Beats GPT-4o & Gemini 2.0 Flash, competitive with DeepSeek v3 at half the active parameters > 1417 ELO on LMArena (chat performance). > Optimized for image understanding, reasoning, and multilingual tasks

Llama 4 Scout:

> 17B active parameters, 16 experts, 109B total parameters

> Best-in-class multimodal model for its size, fits on a single H100 GPU (with Int4 quantization)

> 10M token context window

> Outperforms Gemma 3, Gemini 2.0 Flash-Lite, Mistral 3.1 on benchmarks

Architecture & Innovations

> Mixture-of-Experts (MoE):

First natively multimodal Llama models with MoE

> Llama 4 Maverick: 128 experts, shared expert + routed experts for better efficiency.

Native Multimodality & Early Fusion:

> Jointly pre-trained on text, images, video (30T+ tokens, 2x Llama 3)

> MetaCLIP-based vision encoder, optimized for LLM integration

> Supports multi-image inputs (up to 8 tested, 48 pre-trained)

Long Context & iRoPE Architecture:

> 10M token support (Llama 4 Scout)

> Interleaved attention layers (no positional embeddings)

> Temperature-scaled attention for better length generalization

Training Efficiency:

> FP8 precision (390 TFLOPs/GPU on 32K GPUs for Behemoth)

> MetaP technique: Auto-tuning hyperparameters (learning rates, initialization)

Revamped Pipeline:

> Lightweight Supervised Fine-Tuning (SFT) → Online RL → Lightweight DPO

> Hard-prompt filtering (50%+ easy data removed) for better reasoning/coding

> Continuous Online RL: Adaptive filtering for medium/hard prompts

Why would they need to be trusted? It's either happening or not.

the true measurement of wealth is happiness and security of family

#Bitcoin looking resilient.

Even more interesting that he gets everything right except that Bitcoin is the obvious solution, not China.

Maybe cause red is the communist colour.

⚡️🇨🇳 FUN FACT - A reminder that in China, bearish performances are in green and bullish performances in red.

Complete the square