Claude Code prompt cache metrics now live in /cost: hit ratio, misses, re-cached tokens, warm vs cold — and how to use them to cut your AI spend.
OpenClaw 2026.9.1 is the new beta: gateway restart recovery, Codex runtime 0.150.1 and Node 24 LTS installs — what changed, and whether to update.
Claude Code cross session messaging landed in v2.1.248: here’s what shipped, how sessions alert each other, and how to use it in real workflows.
Apodex FrontierAgent is the new open-source agent framework: ReAct and Agent Team modes, Apache 2.0, no Docker needed. Install steps and my take.
Yutori Navigator n2 explained: the 27B computer-use model that drives GUIs, CLIs and code. Benchmarks, API pricing and how to get started fast.
MiniMax H3 Max review: fal’s post-trained video model makes 5-second clips in ~3 seconds. Pricing, the free tier, benchmarks and my honest verdict.
Claude code self hosted environments explained: the Team and Enterprise public beta running cloud sessions on your own infrastructure, runners and setup.
Deepseek off-peak pricing explained: exact peak hours, the flat 50% discount window, the full rate table and how to time your API workloads.
Qwen 3.8 flash next benchmarks from the official model card: SWE-bench Pro, agentic and multimodal scores, and what the 6B-active MoE means.
What is Claudeforce? Salesforce and Anthropic’s new partnership puts a 37-skill Salesforce plugin inside Claude, with an open beta due September 2026.
Grok 4.6 Microsoft Foundry access is live in public preview: $2 input and $6 output per 1M tokens with managed endpoints. Here’s the full breakdown.
Claude code restricted mode is the new –restricted flag in v2.1.248 that locks Claude Code down to safe file-only tool use. Here’s how it works.
