RTK reports token savings, but our cost benchmarks disagree(quesma.com)

161 points by michalwarda 1 month ago | 80 comments

Token growth indicates future AI spend per dev(blog.kilocode.ai)

167 points by twapi 1 year ago | 132 comments

OpenTelemetry Tracing in < 200 lines of code(jeremymorrell.dev)

213 points by masterj 2 years ago | 28 comments

Once Claude can measure something, it can make it faster(claude.dev)

179 points by matthieu_bl 1 month ago | 113 comments

Engineering blog post from Anthropic's Claude team describing how they made claude.ai three times faster in two weeks by having Claude measure and optimize performance.

How the Claude Code team does code review(claude.dev)

407 points by pretext 10 days ago | 386 comments

An Anthropic engineering blog post by Addy Osmani describing how the Claude Code team reviews the high volume of code produced by AI agents while minimizing human review time.

The real prices of frontier models(playcode.io)

145 points by ianberdin 3 months ago | 72 comments

Measuring Claude 4.7's tokenizer costs(claudecodecamp.com)

678 points by aray07 6 months ago | 474 comments

Blog post measuring Claude Opus 4.7's new tokenizer, finding about 1.47x more tokens than 4.6 on real content versus the documented 1.0–1.35x, and the resulting cost impact.

Claude's system prompt is over 24k tokens with tools(github.com)

584 points by mike210 1 year ago | 308 comments

A GitHub file in a repository of leaked LLM system prompts containing what is claimed to be Claude's full system prompt including tool definitions.

100K Context Windows(anthropic.com)

853 points by samwillis 3 years ago | 364 comments

Anthropic announcement from May 11, 2023 introducing 100K token context windows for its Claude language model.

MCP server that reduces Claude Code context consumption by 98%(mksg.lu)

524 points by mksglu 8 months ago | 100 comments

Blog post introducing Context Mode, an open-source MCP server that sits between Claude Code and tool outputs to cut context window consumption by about 98%.

Claude Code is steganographically marking requests(thereallo.dev)

2318 points by kirushik 3 months ago | 692 comments

Blog post reporting that inspecting Claude Code revealed hidden system prompt markers that vary based on the API base URL and the user's timezone.

Who owns the code Claude Code wrote?(legallayer.substack.com)

504 points by senaevren 6 months ago | 464 comments

A Substack post explaining copyright ownership of AI-generated code from agentic tools like Claude Code, advising builders to document human creative contributions.