Blog
Notes from the engine room.
Product updates, engineering deep-dives, and lessons from building production software with AI.
Airbuild 1.5: Sonnet 5.5, Haiku 5.5 and Mistral Large 4
Sonnet 5.5 and Haiku 5.5 take over as the defaults, Mistral Large 4 joins the picker, and every deployment target gets its own database.
Read more →Airbuild 1.4: Two projects at once, and models that pace themselves
A running turn no longer locks your account, the Claude 5 models choose their own thinking effort, and the workspace speaks your language.
Read more →Airbuild 1.3: A New Name, and Work That Resumes
Version 1.3 pauses a turn when AI Tasks run out instead of failing it, adds three new models and publishes each target from its own branch.
Read more →Hello, Airbuild
We changed our name: XAIO is now Airbuild. Same product, same team, same prices, same projects — and nothing for you to migrate.
Read more →Airbuild 1.2: Prototypes and Website Capture
Version 1.2 brings prototypes the AI sketches before it builds, website capture from a URL, a publish dialog that names the blocker, and budgets.
Read more →How to Build a Website with AI — Step by Step
A practical walkthrough: from a plain-language prompt to a published website with tests and quality checks — in one afternoon, no code required.
Read more →AI Prototype vs. Production: The 10-Point Check
How to tell whether AI-generated software can actually ship: ten concrete checks for architecture, tests, security, ownership and operations.
Read more →What EU Data Residency Means for AI Development
GDPR-native is more than a hosting region: where your code, prompts and data live when you build with AI — and the questions to ask any vendor.
Read more →Airbuild 1.1: Claude 5, Organizations, EU AI Act
Version 1.1 brings Claude 5, Gemini 3.6 and GPT-5.6, the Platform Builder with organizations, project rules, transparent release checks and AI labeling.
Read more →The EU AI Act for Software Teams: What Changed
Since 2 August 2026 the EU AI Act requires transparency for AI interactions. What that means for AI-built software — and how Airbuild labels AI output.
Read more →From Requirements Document to Running App
Bring a spec, screenshots or legacy code and get working software: how Airbuild analyzes what exists, asks the right questions and builds from a plan.
Read more →How We Test AI-Generated Code
Dozens of automated checks per change: how Airbuild generates tests with the code, runs them on every change and blocks releases that fail.
Read more →Airbuild 1.0 Is Here
Version 1.0: new AI models with image understanding, a test cockpit, security and quality checks before publishing, cloud hosting and domains.
Read more →Why We Gate Every Release
All four Airbuild publish gates explained: tests on a frozen snapshot, Semgrep and Trivy scans, code quality, and the open-endpoints check.
Read more →Self-Healing Code Generation
AI-generated code fails sometimes — the question is who fixes it. How Airbuild catches and repairs build errors automatically while it generates.
Read more →