Sung Kim

Huh! Biden loved Bibi, especially his Secretary of State.

So far, Sam did not publicly embrace Bibi.

I believe both Anthropic and OpenAI supports Israel so...

P.S. The last time I switched platforms was when the Vercel CEO publicly embraced Benjamin Netanyahu, whom I despise with every cell of my body. I migrated to Netlify.

Some of you are canceling your ChatGPT Plus or Pro subscriptions in response to recent events.

I’m not doing that. I use both Claude Code and Codex to get my work done, and I’m satisfied with the results.

It kind of was the point.

It's a whoop.

Compute dense human pose estimation from just wifi: github.com/ruvnet/wifi-...

GitHub - alibaba/OpenSandbox: OpenSandbox is a general-purpose sandbox platform for AI applications, offering multi-language SDKs, unified sandbox APIs, and Docker/Kubernetes runtimes for scenarios li... OpenSandbox is a general-purpose sandbox platform for AI applications, offering multi-language SDKs, unified sandbox APIs, and Docker/Kubernetes runtimes for scenarios like Coding Agents, GUI Agent...

Alibaba's OpenSandbox

It is a general-purpose sandbox platform for AI applications, offering multi-language SDKs, unified sandbox APIs, and Docker/Kubernetes runtimes for scenarios like Coding Agents, GUI Agents, Agent Evaluation, AI Code Execution, and RL Training.

github.com/alibaba/Open...

GitHub - second-state/qwen3_asr_rs: Rust implementation of Qwen3-ASR automatic speech recognition Rust implementation of Qwen3-ASR automatic speech recognition - second-state/qwen3_asr_rs

Rust implementation for Speech-to-Text based on Qwen3 models by Michael Yuan

* Self-contained binary build — no external dependencies
* Uses libtorch on Linux with optional Nvidia GPU support
* Uses MLX on MacOS with Apple GPU/NPU support

github.com/second-state...

Reverse Engineering using AI:

- Netflix: gist.github.com/sshh12/dda3a...
- Slack: gist.github.com/sshh12/4cca8...

Understanding RoPE: From Rotary Embeddings to Context Extension | Notion TL;DR: RoPE encodes position via 2D rotations at geometrically spaced frequencies. The RoPE base sets a hard lower bound on effective context length — beyond it, the model literally prefers random tok...

Understanding RoPE: From Rotary Embeddings to Context Extension by Max Li

"RoPE encodes position via 2D rotations at geometrically spaced frequencies. The RoPE base sets a hard lower bound on effective context length — beyond it, the model literally prefers random tokens over similar ones."

This is coded using AI. He is averaging 70+ commits a day.

GitHub - nullclaw/nullclaw: Fastest, smallest, and fully autonomous AI assistant infrastructure written in Zig Fastest, smallest, and fully autonomous AI assistant infrastructure written in Zig - nullclaw/nullclaw

NullClaw

You know, OpenClaw-ish app written in Zig.

github.com/nullclaw/nul...

GitHub - symbolica-ai/ARC-AGI-3-Agents Contribute to symbolica-ai/ARC-AGI-3-Agents development by creating an account on GitHub.

The code used to solve all 3 publicly available ARC-AGI-3 games.

github.com/symbolica-ai...

Financial Datasets MCP Server

This is a Model Context Protocol (MCP) server that provides access to stock market data from Financial Datasets.

github.com/financial-da...

GitHub - kying18/pi-e: policy with experience policy with experience. Contribute to kying18/pi-e development by creating an account on GitHub.

A small repo to deeply understand behavior cloning, action chunking, transformers, and flow matching for embodied AI.

pi-e: Policy with Experience: github.com/kying18/pi-e

LLM-based Evolution as a Universal Optimizer Key take-aways LLM-driven evolution is an efficient, general method for code and agent optimization. We’ve already used code evolution in…

Blog: imbue.com/research/202...
Repo: github.com/imbue-ai/dar...

Evolver, a near-universal optimizer for code and text.

It achieved SOTA (95%) on ARC-AGI-2 (last week that is 😆) and 3x’d performance of the best open model, reaching GPT-5.2-level performance.

Fast LLM Inference From Scratch

Pushing single-GPU inference throughput to the edge without libraries

andrewkchan.dev/posts/yalm.h...

FlashSampling: Fast and Memory-Efficient Exact Sampling

Paper: flashsampling.github.io/FlashSamplin...

3. PyTorch-inspired modeling API with eager-style development + export/lowering to a shared IR for execution and verification

1. Executable IEEE-754 floating-point semantics verified tensor abstractions with precise shape/indexing semantics
2. Formally verified autograd system for differentiation of NN programs Proof-checked certification / verification algorithms like CROWN

TorchLean

A fully verified neural network framework in Lean. TorchLean expands the frontier toward verified neural network software and scientific computing. With the recent release of CSlib, they see this as another step toward a fully verified ML stack.

leandojo.org/torchlean.html

I did. I bought 6 4090s few years back. I could not stand the noise and the need to run the power from different circuit so it will not trip the power. Even garage did not help.

I have no comment. None.

I do not have have parallel sessions running at the same time.

Just to be clear: I’m not affiliated with Anthropic. I’m just a customer.

Upgrade to Claude Max. Escape the timed hell, get the runway you actually need, and reclaim your lifestyle. Your health will thank you.