Seattle Skeptics on AI

Tamara Weed, Feb, 2 2026

Generative AI is transforming public services by helping governments answer citizen questions faster, draft better policies, and manage records more efficiently. In 2026, it's moving from pilot projects to real-world impact.

Tamara Weed, Jan, 31 2026

Post-training evaluation gates are essential safety checks that prevent large language models from deploying with dangerous or broken behavior. Learn how top AI teams use automated and human evaluations to catch failures before users are affected.

Tamara Weed, Jan, 30 2026

A practical guide to building a working generative AI architecture focused on data quality, orchestration, and feedback loops-not just big models. Learn what actually works in enterprise settings and how to avoid the most common failures.

Tamara Weed, Jan, 29 2026

Model access controls determine who can use which LLMs and what they can ask for. Without them, companies risk data leaks, compliance violations, and security breaches. Learn how RBAC, CBAC, and AI guardrails protect sensitive information.

Tamara Weed, Jan, 28 2026

California leads U.S. states in generative AI regulation with strict transparency, consent, and disclosure laws. Colorado, Illinois, and Utah have narrower rules focused on insurance, deepfakes, and privacy-leaving businesses navigating a patchwork of state requirements.

Tamara Weed, Jan, 27 2026

Privacy-Aware RAG protects sensitive data in AI systems by filtering out personal information before it reaches large language models. Learn how it works, where it’s used, and why it’s becoming essential for compliance.

Tamara Weed, Jan, 26 2026

Structured Reasoning Modules transform how LLMs solve complex problems by breaking reasoning into Generate-Verify-Revise steps. This new approach boosts accuracy by over 12% on hard tasks and reduces errors, making it essential for finance, science, and legal AI systems.

Tamara Weed, Jan, 25 2026

Decoder-only transformers dominate modern LLMs for speed and scalability, but encoder-decoder models still lead in precision tasks like translation and summarization. Learn which architecture fits your use case in 2026.

Tamara Weed, Jan, 24 2026

Prompt chaining and agentic planning are two ways to make LLMs handle complex tasks. One is simple and cheap. The other is smart but costly. Learn which one fits your use case-and why most teams get it wrong.

Tamara Weed, Jan, 23 2026

Agentic LLMs plan, use tools, and act autonomously-transforming AI from passive responders to active problem-solvers. Learn how they work, where they're used, and why safety remains a critical challenge.

Tamara Weed, Jan, 22 2026

Vibe coding tools generate code fast but fail at system design. Today's platforms can build components but not scalable architectures. The next wave must solve context, governance, and planning gaps to move beyond prototypes.

Tamara Weed, Jan, 21 2026

Vibe coding lets teams build software by describing what they want-no code editing needed. Learn the five rules, the right tools, and when to use it-plus the risks of skipping tests and reviews.