Code-A1: Adversarial Co-Evolution Eliminates Test Suite Dependency.

Shared entity: Code / Same source domain / Shared topic / What happened next / Tension

RSIBench-Data: agents improve on their first attempt 58% of the time but finish below their own peak in 78% of runs.

Both cover Code; reported by the same outlet (arxiv.org); overlapping topics (agent, benchmark, code).

Engram: a lean retrieved context beats stuffing the full history into the window.

Both cover Code; reported by the same outlet (arxiv.org); overlapping topics (agent, benchmark, code).

Shared entity: Code / Same source domain / Shared topic / What happened next

Airbnb says AI writes 60% of its new code and cut concept-to-launch by 60%

Both cover Code; reported by the same outlet (arxiv.org); overlapping topics (agent, benchmark, code, coding).

Shared entity: Code / Shared topic / What happened next / Tension

Microsoft's MAI coding models reach Copilot and VS Code.

Both cover Code; overlapping topics (benchmark, code, coding, model); picks up the Code thread on 2026-06-07.

GitHub Copilot Switches to Token-Based Billing Today. Developers Report 25x Cost Increases.

Both cover Code; overlapping topics (agent, code, coding, model); picks up the Code thread on 2026-06-01.

Shared entity: Code / Same source domain / Shared topic / What happened next

Mendel Gödel Machine borrows mutation operators from biology so self-rewriting agents learn from many trajectories, not one failure.

Both cover Code; reported by the same outlet (arxiv.org); overlapping topics (agent, bottleneck, code).

A survey of 85 agentic-security papers finds attack work outpaces defense 3.9:1, and the action layer gets 4.7% of attention.

Both cover Code; reported by the same outlet (arxiv.org); overlapping topics (adversarial, agent, code).

Tencent: poisoned agent experience becomes a permanent skill

Both cover Code; reported by the same outlet (arxiv.org); overlapping topics (agent, become, code).