Morning · Tuesday, June 9

AI Brief

Self-evolving agents with acceptance tests; stealable pattern for Claude Code.

  • 12 items
  • 2 axes
  • 3 sources
  • Generated 12:12 UTC

TL;DR

  1. 01
    rss/arxiv-cs-ai Self-evolving agents with acceptance tests; stealable pattern for Claude Code.
  2. 02
    rss/arxiv-cs-ai Durable write surface for LLM coding agents via MCP
  3. 03
    hn Extend Claude Code session length

Business 1 item · Roux outbound · pitch sites · agency ops

  1. 01

    DeepSeek enters the fight for token volume, Anthropic continues to dominate spend

    rss/vercel-blog yesterday score 8

    Token spend data informs provider choice for Roux

    vercel.com/blog

Skills 11 items · Claude Code patterns worth stealing

  1. 01

    PACE: Anytime-Valid Acceptance Tests for Self-Evolving Agents

    rss/arxiv-cs-ai 8h ago score 9

    Self-evolving agents with acceptance tests; stealable pattern for Claude Code.

    arxiv.org/abs
  2. 02

    Resilient Write: A Six-Layer Durable Write Surface for LLM Coding Agents

    rss/arxiv-cs-ai 8h ago score 9

    Durable write surface for LLM coding agents via MCP

    arxiv.org/abs
  3. 03

    How to have Claude Code run 28× longer before auto-compaction

    hn 10h ago score 8

    Extend Claude Code session length

    github.com/wiztek-llc
  4. 04

    affaan-m/ECC ★211,423

    github 9h ago score 8

    Claude Code performance optimization system with skills/instincts.

    github.com/affaan-m
  5. 05

    SKILL.nb: Selective Formalization and Gated Execution for Durable Agent Workflows

    rss/arxiv-cs-ai 8h ago score 8

    Durable agent workflows pattern; directly usable for Claude Code skills.

    arxiv.org/abs
  6. 06

    Context Rot in AI-Assisted Software Development: Repurposing Documentation Consistency for AI Configuration Artifacts

    rss/arxiv-cs-ai 8h ago score 8

    Directly addresses AI config artifacts like CLAUDE.md

    arxiv.org/abs
  7. 07

    Efficient and Scalable Provenance Tracking for LLM-Generated Code Snippets

    rss/arxiv-cs-ai 8h ago score 8

    Provenance tracking for LLM code, actionable pattern.

    arxiv.org/abs
  8. 08

    Show HN: Built an open-source local firewall for AI coding agents

    hn 7h ago score 7

    Open-source firewall for coding agents

    news.ycombinator.com/item
  9. 09

    Egonex-AI/Understand-Anything ★55,672

    github 6h ago score 7

    Interactive knowledge graph tool for codebases, works with Claude Code.

    github.com/Egonex-AI
  10. 10

    obra/superpowers ★222,001

    github 6h ago score 7

    Agentic skills framework for Claude Code and other AI tools.

    github.com/obra
  11. 11

    How an Agent Built a 3D Paris Gallery by Chaining Two Hugging Face Spaces

    rss/huggingface 1h ago score 7

    Agent chaining pattern for Claude Code workflows

    huggingface.co/blog