Paper Clips

Read Less, Know More

Saturday, February 21, 2026
Simon Willison

Quoting Thibault Sottiaux

OpenAI has improved GPT-5.3-Codex-Spark performance by 30%, now serving at over 1200 tokens per second. This represents a significant speed improvement for AI code generation capabilities. <div...

Simon Willison

Andrej Karpathy talks about "Claws"

Andrej Karpathy discusses "Claws," a new category of AI agent systems that he believes represents the next evolution in AI orchestration. He views Claws as a new layer built on top of LLM agents,...

Friday, February 20, 2026
anthropics/claude-code

anthropics/claude-code v2.1.50

This release focuses heavily on memory leak fixes and performance improvements for long-running sessions. It also adds agent worktree isolation capabilities and expands the 1M context window to Opus...

Simon Willison

Adding TILs, releases, museums, tools and research to my blog

Simon Willison has added a new "beats" feature to his blog that consolidates his online activity from multiple platforms into a single timeline. This integration pulls in content from five different...

Simon Willison

Taalas serves Llama 3.1 8B at 17,000 tokens/second

Canadian startup Taalas has developed custom hardware that runs Llama 3.1 8B at an unprecedented 17,000 tokens per second. Their breakthrough speed makes AI responses nearly instantaneous,...

Simon Willison

ggml.ai joins Hugging Face to ensure the long-term progress of Local AI

Hugging Face has acquired ggml.ai, the team behind llama.cpp - the breakthrough tool that made running large language models on consumer hardware possible. This acquisition could accelerate...

Peter H. Diamandis

Tony Robbins Warned President Obama About AI Job Loss

Tony Robbins discusses his warning to President Obama about AI's potential impact on job displacement. The conversation explores how AI will fundamentally transform the labor market and the need for...

AI News & Strategy Daily | Nate B Jones

The Job That Replaced Senior Developers Costs $20K/Month (More Output with Fewer Engineers)

The fundamental unit of computing work is shifting from instructions to tokens, creating a new economic model where intelligence becomes a purchasable commodity. This transformation is reshaping...

Matt Ganzak

You should own your IP! Thanks to #openclaw for helping me realize this.

Overview

Swift.org

Announcing Swift System Metrics 1.0: Process-Level Monitoring

Apple has released Swift System Metrics 1.0, a Swift package that collects process-level system metrics like CPU and memory usage across Linux and macOS. This enables developers to monitor...