news

Claude Code’s product lead talks usage limits, transparency, and the “lean harness” - Ars Technica

Ars interviews Claude Code product head Cat Wu after Anthropic doubled usage limits: idle-broken caches make the next query far more expensive, and some plugins silently spawn around a hundred subagents per run.

Published 2026-05-15Source: Ars Technica

Why it matters

This is the most concrete public account of agentic token waste from the team that debugs it — live, on calls with users, from locally stored transcripts. Dario Amodei's framing explains the squeeze: Anthropic planned for 10x annual growth and got 80x.

Tokenmaxxing read

Wu's two waste patterns are checkable today: /usage flags broken-cache sessions and heavy subagent fan-out — audit both before blaming the model for quota burn. The team's stated north star is intelligence per token.

Source takeaway

Token efficiency is managed at the harness layer, not the model layer: Anthropic defaults to not shipping a tool unless it clearly improves token performance or accuracy.

Topic links

agentstopic coding-agentstopicscoreboardstoken-consumptiontopictokenmaxxing

Related projects

Tools that match this angle

#4In spirit

Agents

LangGraph

langchain-ai/langgraph

A framework for building resilient stateful agents with explicit graphs, persistence, human-in-the-loop flows, and controllable execution.

34.8K5.8KMIT

agentsstateworkflows

Project profile GitHub

#5Direct

Evaluation

promptfoo

promptfoo/promptfoo

A CLI and CI workflow for testing prompts, agents, and RAG systems across models, with evals and red-team style checks.

22.2K2KMIT

prompt-evalscirag

Project profile GitHub

#6In spirit

Evaluation

DSPy

stanfordnlp/dspy

A framework for programming and optimizing language-model pipelines rather than hand-tuning one prompt at a time.

35K3KMIT

optimizationprogrammingevals

Project profile GitHub

Related feed

More source-linked context

newsA

news2026-06-09

Claude Fable 5 and Claude Mythos 5 - Anthropic

Anthropic shipped Claude Fable 5 (GA, with classifier safeguards) and Claude Mythos 5 (safeguards lifted, vetted partners only) on June 9 — $10 per million input tokens, $50 per million output, under half the Mythos Preview price.

agentscoding-agentspricing

Read note

newsYF

news2026-05-24

AI Cost Crisis Emerges as Claude Usage and Agentic Coding Bills Spiral

Yahoo Finance flags an emerging AI cost crunch: agentic coding and heavy Claude usage can spike bills fast, forcing leaders to rethink budgeting and ROI.

tokenmaxxingcoding-agentsagents

Read note

newsHN

news2026-05-14

Microsoft’s WinUI agent plugin trims token use by over 70% during development - Help Net Security

Help Net Security covers Microsoft's WinUI agent plugin for GitHub Copilot CLI and Claude Code, aiming to make WinUI 3 app loops (build/run/test/package) agent-friendly.

tokenmaxxingcoding-agentsagents

Read note