Notebook
Blog
Insights, tutorials, and news about web development.
The Google Workspace security dashboard: roles, detection, and emailThe Google Workspace security dashboard: roles, detection, and email
The complete walkthrough of Workspace security: least privilege and Context-Aware Access, Security Center with activity rules, SAML and scopes, Chrome and devices, and SPF/DKIM/DMARC in both directions of email.
From waveform to text and back: how AI voice respondsFrom waveform to text and back: how AI voice responds
The two pipelines of AI voice: recognition in six stages with MFCC in the middle, synthesis in four stages with G2P and prosody, and where each is already used.
From pixel to description: how a machine seesFrom pixel to description: how a machine sees
How a machine sees: the image as a matrix of numbers, the four computer vision tasks, the CNN versus the vision transformer with attention, and generation by diffusion working backwards.
Networks for AI on Google Cloud: goodput, the dual plane, and model-aware routingNetworks for AI on Google Cloud: goodput, the dual plane, and model-aware routing
The three phases of the AI network on Google Cloud: goodput as the metric, rail-aligned GPUs with a dual traffic plane, massive ingestion, multi-cloud, and why round-robin fell short for inference.
Security in the world of AI: the real risks, the four components, and the SAIF frameworkSecurity in the world of AI: the real risks, the four components, and the SAIF framework
The real risks of AI in the company, the four components of any AI system that must be secured, and SAIF: four steps and six elements of Google's framework for securing AI systems.
Prioritize risks with a simulated attacker: Security Command Center and its digital twinPrioritize risks with a simulated attacker: Security Command Center and its digital twin
Virtual red teaming against a digital twin of the environment: declare high-value resources, read chokepoints and toxic combinations, and always remediate in order of impact.