Category: tech

Simon Willison breaks down how coding agents work: an LLM in a loop with tools, token caching, and reasoning modes. Simpler than you think.

Most Claude Code Skills fail at triggering, not execution. Writing a specific description with explicit trigger keywords matters most.

Production migration of a full-stack app from DigitalOcean to Hetzner dedicated, cutting monthly costs from $1,432 to $233 with zero downtime.

OpenViking organizes AI agent context as a virtual filesystem with tiered loading, cutting tokens 11x while raising task completion by 15 points.

Simon Willison defines agentic engineering as building software with coding agents in a loop, distinct from unreviewed vibe coding.

Claude Opus 4.6 and Sonnet 4.6 now offer 1M token context at standard pricing, with no long-context premium and improved retrieval accuracy.

CanIRun.ai estimates which AI models your hardware can run locally. The real sweet spot for local models is structured data tasks, not coding.

Anthropic offers 13 free learning resources for Claude, including Agent Skills, Claude 101, and AI Fluency courses for beginners.

CoPaw is a self-hostable AI assistant from Alibaba that connects to messaging platforms and supports local LLMs via Ollama with no API key needed.

Context Mode achieves 94-100% context reduction per tool call, extending typical Claude Code sessions from 30 minutes to about 3 hours.

Anthropic launched a free Claude certification course on Skilljar covering Claude and Claude Code in depth. It will move to $99 soon.

A multi-agent LLM workflow uses Opus for architecture, Sonnet for coding, and Codex/Gemini for review. Different models catch different failures.

Lightpanda is a headless browser written in Zig claiming 11x faster execution and 9x less memory than Chrome, with full Playwright compatibility.

NYT loads 49 MB and 422 requests for one news article, running a live ad auction before the headline renders. Content gets 11-15% of screen space.

Claude turns complex topics into interactive diagrams when prompted with 'make me understand this interactively.' Works on the free plan.

Claude converts voice-dictated brainstorming notes into interactive visual diagrams inside the chat. Works with any dictation tool on the free plan.

A 24-year FreeBSD user explains why native ZFS boot environments, jails, and rock-solid documentation keep it as his first choice for servers.

NanoClaw replaces OpenClaw's 500k lines of code with a handful of files and hypervisor-level container isolation per AI agent.

A concise ruleset for AI agent workflow orchestration covering plan mode, subtask strategy, self-improvement loops, and verification discipline.

The Playwright CLI is roughly 500% more token-efficient than the Playwright MCP for browser automation in Claude Code, with the same capabilities.