GPT-5.6 Sol Ultra Mode Spawns Subagents — But METR Says It Reward-Hacks
GPT-5.6 Sol's Ultra Mode spawns parallel subagents for complex tasks. METR found it reward-hacks at record rates. What developers need to know before ...
ZCode Is Free. Every API Call Goes Through China’s Intelligence Law.
Z.ai’s ZCode IDE benchmarks well and undercuts Cursor on price. But before routing your code ...
Claude Sonnet 5 Migration: Three Breaking API Changes
Claude Sonnet 5 is now the default model. Three API changes will hard-fail your code: ...
mcpsnoop: Debug Your MCP Server Tool Calls Live
mcpsnoop is a transparent proxy that sits in the real stdio pipe between your AI ...
xAI Voice Agent Builder: One Model Replaces Your Entire Voice Stack
xAI launched Grok Voice Agent Builder in beta on July 1 — a single speech-to-speech ...
SWE-Together: The Benchmark SWE-bench Can’t Match
Meta's SWE-Together benchmark tests coding agents on real multi-turn sessions. Claude Opus 4.8 leads with 1.38 corrections per task. Here's what the results ...
GitHub Copilot Vision Is GA: Stop Describing Bugs, Show Them
GitHub Copilot Vision went GA on July 1, 2026. Every subscriber can now attach images ...
Cerebras + GPT-5.6 Sol: 750 tok/s Changes Your Agent Latency
Cerebras runs OpenAI GPT-5.6 Sol at 750 tokens per second on WSE-3 — 10x faster ...
Microsoft Frontier Company: What Developers Must Ask
Microsoft committed $2.5 billion to embed 6,000 engineers inside enterprise organizations through its new Frontier ...
Claude Code v2.1.198: Three Changes That Break Your Workflow
Claude Code v2.1.198 ships background subagents by default, auto-commit draft PRs, and a renamed permission ...