Thomas Wiegold

Blog

Insights on AI solutions, web development, and modern software engineering practices.

19 articles published

code

How to Build an MCP Server with TypeScript

Build an MCP server with TypeScript using the official SDK. Covers tools, resources, error handling, and production deployment for AI integrations.

ClaudeMCPTypeScript+2
Read more
code

Building Reliable Invoice Extraction Prompts That Handle Edge Cases

Learn how to craft LLM prompts for invoice extraction that handle messy scans, edge cases, and human error—with confidence signals and real-world testing strategies.

ClaudeLLMTypeScript+2
Read more
AI Solutions

AI for Small Business: 7 Use Cases That Actually Work

55% of small businesses now use AI, but most don't know where to start. Here are the proven use cases delivering real ROI—no massive budget required.

SMBBusinessAI
Read more
AI Solutions

Do LLMs Actually Understand Code? The Evidence

Do LLMs truly understand code or just match patterns? 2024-25 research reveals a surprising paradox—and what it means for your development workflow.

LLMCodeResearch
Read more
AI Solutions

Gemini 3 Flash: Why Google's Budget Model Is My New Default

Google's Gemini 3 Flash beats GPT-5.2 benchmarks at 3.5x lower cost. Here's why this workhorse model changes the economics of AI in production apps.

GoogleGeminiFlash+1
Read more
AI Solutions

JetBrains Shuts Down Fleet, Bets Everything on Agentic AI with Air

JetBrains kills Fleet after 4 years in preview. Its replacement, Air, wants you to stop coding and start supervising AI agents instead. Here's what happened.

JetbrainsAirFleet+1
Read more
code

Claude API Memory Tool: Build Agents That Learn

Claude's Memory Tool lets your AI agents retain knowledge across sessions. Learn how to implement persistent memory without building your own solution.

ClaudeAPIMemory+1
Read more
AI Solutions

Claude Opus 4.5 Review: Anthropic's New Coding Model Breaks Records

Claude Opus 4.5 achieves 80.9% on SWE-bench with 67% lower costs. Hands-on review of the new effort parameter, token efficiency, and real coding performance.

AnthropicClaudeOpus+1
Read more
AI Solutions

Google Gemini 3 Hits #1 on LMArena: A Developer's Honest First Impressions

Gemini 3 Pro hits #1 on LMArena with 1501 Elo. A developer's honest first impressions and testing plan vs Claude Sonnet 4.5 for real coding work.

GoogleGeminiLLM
Read more