Blog

Engineering notes on LLM inference costs, GPU spot pricing, IDE security, and building developer tools — from the AIVory team.

The GPU Price Crash of 2025 and Why It Reversed

H100 rental prices fell 80% from 2023 to 2025, then reversed 40% in 2026. Full pricing data across providers.

The Minimal AWS Stack for an AI Product

VPC, two subnets, one EC2 instance, one RDS database, an S3 bucket, and a load balancer. That is the entire stack.

AIVory Guard vs Semgrep: Scanning AI-Generated Code

Semgrep is open source, mature, and free for most teams. Here is how AIVory Guard's IDE-native compliance scanning actually differs, checked point by point.

Can You Run a Production LLM on Spot GPUs

Spot GPUs cost 60-70% less than on-demand. The catch is they get reclaimed mid-inference. Here is how to architect around that.

AIVory Architect vs Brainboard: AI Terraform Compared

Brainboard is a mature, team-built cloud design platform. Here is how AIVory Architect's IDE-native canvas actually compares, feature by feature.

I Cut My AI Inference and GPU Costs in Half With One Line of Code

I built an AI product and forgot about the inference costs. For about a month. Then I built a router.

Securing AI Code: OWASP, GDPR, HIPAA, EU AI Act Checklist

A concrete, bookmarkable checklist for AI-generated code against OWASP Top 10, GDPR, HIPAA, and the EU AI Act, with the actual check for each item.

LiteLLM vs Smart Inference: Self-Hosted or Managed

A concrete look at what you gain and lose self-hosting LiteLLM versus using a managed router like Smart Inference, for teams routing open-weight models.

What Is a GPU Spot Instance? Explained Simply

A GPU spot instance is spare cloud GPU capacity sold at a steep discount. How spot pricing works, the interruption trade-off, and when to use it.

What Is an LLM Router? A Plain-English Guide

An LLM router sits between your app and multiple model providers, picking the best one per request. How it works, when you need one, and the trade-offs.