Articles by Grizzly Peak Software

I pay for four AI subscriptions and tested nine more. Here's which ones earn their monthly fee and which ones you can skip.

Mistral Pro is $14.99/mo with a 150-message soft cap. Here's what the free tier actually gives you and what the API costs per token.

I replaced my inference backend with DigitalOcean's serverless API and GPT-oss-120b. Here's how I wired it into a Node.js app and what I learned.

Claude, ChatGPT, and Gemini all handle usage limits differently now. Here's what each tier actually gives you and when you'll hit the wall.

Groq API free tier limits in 2026: exact RPM, RPD, TPM, and TPD for every model including Llama 3.3, Llama 4 Scout, and Whisper.

Two malicious LiteLLM versions hit PyPI for a few hours. They tried to steal your API keys. Here's what happened and what to do about it.

I tested Skywork's agentic AI workspace. It one-shotted an SEO audit of my site. Here's the honest review and what engineers should know.

Most "agentic AI" content is hand-wavy architecture diagrams. This covers the agent loop, tool design, memory patterns, and MCP with working code.

Skip the PhD. A working developer's roadmap to AI in 2026: LLMs, agents, MCP, RAG, and the projects that actually build real skills.

OpenClaw has 247K stars. Its creator started with a simple agent loop. You should too. Build the fundamentals before adopting the framework.

My AI publishing pipeline got a book flagged on Amazon. Here's how I added content risk assessment and why every agentic workflow needs quality gates.

Rome's road network, provincial governance, and currency debasement mirror the same failures we build into software. The parallels are uncomfortable.

Five-layer guardrail architecture for AI document extraction pipelines that fabricate data instead of failing visibly.

AI delivery pipelines need governance for accountability and compliance. Add traceability, approval gates, and audit trails without crushing velocity.

After 40+ years in software, here are the engineering skills AI has devalued and the ones worth doubling down on right now.

I built an AI chatbot that diagnoses car problems and generates repair content. Here's what I learned about trust, accuracy, and domain-specific AI.

80+ books, 70 audiobooks, one AI-assisted pipeline. Here's why the "AI content is slop" argument confuses the tool with the operator.

Your AI writes fast. It also writes JWT bugs, hardcoded secrets, and SQL injection. Here's what the scanners actually found.

Six months of tracked hours and output reveal where AI tooling actually boosts solo developer productivity and where it quietly costs you time.

The agent pattern that made Cursor powerful is moving into docs, spreadsheets, and email. The engineering challenges are harder than you think.