Anthropic accelerated claude.ai and the Claude desktop app by 3x in two weeks using Claude itself to identify performance bottlenecks and ship over 3,000 changes. The team focused on four high-impact user journeys, reducing page load times from 3.1 to 0.55 seconds and new session starts from 0.8 to 0.3 seconds, saving tens of thousands of user-hours daily. Claude Tag (beta) autonomously monitored deploys, built benchmarks, and proposed improvements while humans steered through goal-setting and change approval.
An AI software factory is a managed system that converts defined requirements into validated software changes using AI agents, tools, automation, and human oversight. It differs from coding agents by connecting business intent through context, execution, verification, approval, release, and feedback loops. Camplight has deployed three client implementations and created Nest, a public generalized model based on their experience.
A comparison of three AI models (GPT-5.6-Terra, DeepSeek-V4-Flash, and Claude Sonnet 5) tested their ability to generate Ansible configurations across three scenarios of increasing complexity. Spotter analysis found 140 errors, 94 warnings, and 221 hints total, with missing fully qualified module names being the most common error type across all models.
Actwise Intel publishes a Top Idea for a deployment planning system that maps AI workloads to regional and sovereignty requirements while producing auditable deployment records. The idea scored 79/100 (99th percentile) and is available under CC BY 4.0 for commercial use with attribution.