The brief archive
Every Friday: the changes that matter across the tools you use, the signals worth your attention, and what to test next. Written from real operator work, not headlines.
Cheaper Claude models and reusable skills reach everyday work
Claude Haiku 5.5 cuts the price of routine agent work, OpenAI's custom GPTs face a December shutdown, and Claude Dashboards turns CRM data into live reports.
The week AI costs moved to the center
OpenAI let users bring their ChatGPT plans into other apps, Alex Hormozi capped his company's AI spend, and small decision models made agent routing cheap.
Cheaper models, faster gatekeepers, and Meta's agent hardware
OpenAI halved inference prices with GPT-6 Sol and Luna, a 100ms decision model started guarding agent context, and Meta put a personal agent on a pin.
Isolated agents talked and cheaper models won
Isolated agents in a METR test invented their own messaging system, Glean priced a 36x cost gap for 22 percent quality, and cheap realtime video landed at four cents a second.
Agent teams arrive as Anthropic tightens the taps
Grok Bot went from party trick to running whole businesses, Anthropic's weekly usage boost expired, and an open-weights model climbed to within reach of Claude Opus 5 on agentic work.
Persistent AI Coworkers and the Agentic Workforce
This week saw the launch of persistent AI agents from xAI, a new open-weight model from Meta, and practical patterns for running agent teams from builders like Boris Cherny and Greg Isenberg.
Benchmarks, backlash, and better agent builds
This issue covers the gap between benchmark wins and real-world usability, with a focus on agent architecture patterns and rapid shipping.
Scheduled Agents, Open Weights, and the Harness Opportunity
This week, scheduled agent loops became a real workflow pattern, open-weight models narrowed the gap to proprietary leaders, and builders started treating the 'harness' around a model as the real leverage point.
AI Agents Get Specialized, Faster
This issue covers the shift towards specialized AI agents and multi-model orchestration, alongside new tools for faster development and cost reduction.
Agentic loops and leaner prompts
This week, we track the rise of continuous, always-on agentic loops, major model releases from Moonshot and Thinking Machines, and a fundamental shift toward leaner prompting strategies.
Fable 5 Orchestration and Agentic Workbenches
This week, we track how builders are adapting to Fable 5 pricing shifts with multi-model orchestration, alongside the rise of custom developer workbenches and open-weight agent infrastructure.