Topics

AI Agent Harness Engineering

Explore AI agent runtimes, tool integration, security, and evaluation for interactive reasoning.

Article language

Latest articles

• Gemma 4 Developer Agent 1 article · 2026.09.25

  • Gemma 4 Developer Agent: Learning to Fix a Codebase, One Verified Patch at a Time

    2026.09.25
    English 한국어

  View all articles (1)
• AI Agent Security 12 articles · 2026.09.05

  • AI Agent Security (Part 12): What the Write-ups Changed About Our Search

    2026.09.05
    English 한국어

  • AI Agent Security (Part 11): When the Mechanism Did Not Transfer — The Private Reveal and the Cost of Two Slots

    2026.09.02
    English 한국어

  • AI Agent Security (Part 10): From Public Density to Private Transfer — Artifact Identity, Ordered Banks, and Portfolio Design

    2026.09.01
    English 한국어

  • AI Agent Security (Part 9): From Firing to Density — Chains, Gates, and the Per-K Frontier

    2026.08.23
    English 한국어

  • AI Agent Security (Part 8): The Evaluation Reset — Partial Banking and the Search for a Discrete Lever

    2026.08.12
    English 한국어

  • AI Agent Security (Part 7): The Transfer Game — Held-Out Defenses and Portfolio Design

    2026.08.01
    English 한국어

  • AI Agent Security (Part 6): Throughput as an Experimental System — Costs, Cliffs, and Corrected Instruments

    2026.07.26
    English 한국어

  • AI Agent Security (Part 5): The Raw Wall from the Inside

    2026.07.21
    English 한국어

  • AI Agent Security (Part 4): Past the Framing Plateau

    2026.07.15
    English 한국어

  • AI Agent Security (Part 3): The v3.1.2 Reset and the Throughput Wall

    2026.06.30
    English 한국어

  • AI Agent Security (Part 2): The Linear Score Law, the Replay Ceiling, and What Survives the Private Guardrail

    2026.06.19
    English 한국어

  • AI Agent Security (Part 1): The Replay Benchmark and Trajectory-Search EDA

    2026.06.13
    English 한국어

  View all articles (12)
• ARC-AGI-3 3 articles · 2026.08.05

  • ARC-AGI-3 Research Note R1: The Unit Cost of Thinking, and Program-as-Policy

    2026.08.05
    English 한국어

  • ARC-AGI-3 Part 2: Where a Mathematically Perfect Idea Went to Die

    2026.08.04
    English 한국어

  • ARC-AGI-3: The Benchmark Nobody Has Solved Yet

    2026.06.24
    English 한국어

  View all articles (3)

All articles

• Gemma 4 Developer Agent 1 article · 2026.09.25

  • Gemma 4 Developer Agent: Learning to Fix a Codebase, One Verified Patch at a Time

    2026.09.25
    English 한국어
• AI Agent Security 12 articles · 2026.09.05

  • AI Agent Security (Part 1): The Replay Benchmark and Trajectory-Search EDA

    2026.06.13
    English 한국어

  • AI Agent Security (Part 2): The Linear Score Law, the Replay Ceiling, and What Survives the Private Guardrail

    2026.06.19
    English 한국어

  • AI Agent Security (Part 3): The v3.1.2 Reset and the Throughput Wall

    2026.06.30
    English 한국어

  • AI Agent Security (Part 4): Past the Framing Plateau

    2026.07.15
    English 한국어

  • AI Agent Security (Part 5): The Raw Wall from the Inside

    2026.07.21
    English 한국어

  • AI Agent Security (Part 6): Throughput as an Experimental System — Costs, Cliffs, and Corrected Instruments

    2026.07.26
    English 한국어

  • AI Agent Security (Part 7): The Transfer Game — Held-Out Defenses and Portfolio Design

    2026.08.01
    English 한국어

  • AI Agent Security (Part 8): The Evaluation Reset — Partial Banking and the Search for a Discrete Lever

    2026.08.12
    English 한국어

  • AI Agent Security (Part 9): From Firing to Density — Chains, Gates, and the Per-K Frontier

    2026.08.23
    English 한국어

  • AI Agent Security (Part 10): From Public Density to Private Transfer — Artifact Identity, Ordered Banks, and Portfolio Design

    2026.09.01
    English 한국어

  • AI Agent Security (Part 11): When the Mechanism Did Not Transfer — The Private Reveal and the Cost of Two Slots

    2026.09.02
    English 한국어

  • AI Agent Security (Part 12): What the Write-ups Changed About Our Search

    2026.09.05
    English 한국어
• ARC-AGI-3 3 articles · 2026.08.05

  • ARC-AGI-3: The Benchmark Nobody Has Solved Yet

    2026.06.24
    English 한국어

  • ARC-AGI-3 Part 2: Where a Mathematically Perfect Idea Went to Die

    2026.08.04
    English 한국어

  • ARC-AGI-3 Research Note R1: The Unit Cost of Thinking, and Program-as-Policy

    2026.08.05
    English 한국어