Kimi-K3 solved 94% of workflows once. Only 13% survived every run
New evidence separates occasional agent capability from dependable operation, while applied teams show where task choice, model training, and runtime retrieval each belong.
The bigger picture · 9 editions
Weekly Review
A deeper synthesis of models, agents, infrastructure and research. We focus on the patterns that should change architectural assumptions.
October 2026
September 2026
Meta’s agent handed calls to people. Delegation needs its own control plane
Muse exposed a specific agent boundary: authenticating the visible service does not reveal who performed an action, what authority crossed the handoff, or who owes the user a remedy.
AI teams are starting to measure work, not usage
OpenAI and Anthropic exposed richer task-level metrics this week. The architectural opportunity is a task ledger that connects model activity, human intervention, operational effects, and business outcomes without pretending correlation is causation.
AI changed the build cost. Verification now sets the architecture
Shopify, Mistral, OpenAI and independent practitioners showed that faster generation changes system design only when teams make semantics, tests and review executable.
The system frontier moved more than the model frontier
A week of frontier launches made one thing clearer: persistent state, safeguards, evaluation and agent research loops increasingly determine the deployable AI system.
August 2026
An agent’s environment is part of its capability and threat model
The August 24 to 30 review connects emergent coordination, harness optimization, agent-oriented models and infrastructure reliability.
The harness has become part of the capability
The August 17 to 23 review examines runtime safety, agent architecture and the practical costs of operating capable systems.
The AI stack is separating into distinct execution tiers
The August 10 to 16 review connects model portfolios, local agents, planner to executor architectures and the growing role of the harness.
Model capability and system reliability are separating
The August 3 to 9 review connects agent containment, durable runtimes, model routing, evaluation and deployment economics.