# Blog | Nadya Construction Studio

> Practical articles on AI systems, online payments, cloud reviews, cloud and AI cost optimization, and website speed from a New York software studio.

Canonical: https://nadiaconstructionstudio.com/blog

[Home](https://nadiaconstructionstudio.com/) / Blog

Blog

# Notes from *the build.*

Practical, sourced articles on the things we build every day.

[6 min read

### How Much Does It Cost to Build an AI Agent in 2026?

What an AI agent really costs in 2026: build vs run, token math per task, why agents burn more tokens than chat, and how evals, caching and routing cut spend.Read the article](https://nadiaconstructionstudio.com/blog/ai-agent-cost)[5 min read

### Getting Cited by ChatGPT, Claude and Google AI Overviews

What gets a site cited in AI answers, per the evidence: Google's AI features guidance, crawler access, llms.txt data, the GEO paper's limits and Ahrefs studies.Read the article](https://nadiaconstructionstudio.com/blog/ai-search-geo)[5 min read

### What an AWS Review Should Check: Cost, Security, Reliability

A practical AWS review checklist built on the Well-Architected Framework: root and IAM security, tested backups, rightsizing, Savings Plans and hidden costs.Read the article](https://nadiaconstructionstudio.com/blog/aws-review-checklist)[6 min read

### How to Cut Your AWS and AI Bill Without Breaking Anything

A safe order for cutting AWS and LLM costs: visibility, idle resources, rightsizing, Savings Plans, storage and NAT fixes, then caching, batching and limits.Read the article](https://nadiaconstructionstudio.com/blog/cloud-cost-optimization)[5 min read

### Core Web Vitals in 2026: What to Fix First

Core Web Vitals in 2026: the LCP, INP and CLS thresholds, why LCP fails most often on mobile, and the fixes to make first, based on web.dev and Google guidance.Read the article](https://nadiaconstructionstudio.com/blog/core-web-vitals-2026)[6 min read

### Evals First: How to Know Your AI Agent Actually Works

How to evaluate an AI agent before launch: real test tasks, code and model graders, pass@k vs pass^k, regression suites, and reading transcripts that matter.Read the article](https://nadiaconstructionstudio.com/blog/evals-before-agents)[6 min read

### Stripe Webhooks Without Double Charges or Lost Orders

How to make Stripe payments safe to retry: idempotency keys, duplicate webhook events, signature checks, fulfillment that runs once, and daily reconciliation.Read the article](https://nadiaconstructionstudio.com/blog/stripe-webhooks)

From MenteDB

## On AI *memory.*

[From MenteDB

### Infinite Context Window for AI with Memory

How long term memory lets an assistant work past the limits of its context window.Read on mentedb.com](https://mentedb.com/blog/infinite-context-window-for-ai-with-memory)[From MenteDB

### Your agent knows your preferences. It just never uses them

Why stored preferences go unused, and how memory recall fixes it.Read on mentedb.com](https://mentedb.com/blog/your-agent-knows-your-preferences-it-just-never-uses-them)[From MenteDB

### Why your agent keeps repeating facts you already corrected

Contradictions, stale beliefs and how a memory engine should handle corrections.Read on mentedb.com](https://mentedb.com/blog/why-your-agent-keeps-repeating-facts-you-already-corrected)