Batch API 50% Discount: When to Use It at Anthropic, OpenAI and Google
How the Claude, OpenAI and Gemini batch APIs work: the 50% discount, 24-hour windows, size limits, request formats, good-fit jobs and a monthly example.
Practical AI, measured
What AI actually costs, and how to build with it.
ailog is a small, independent notebook. Every number here comes from an official source we link to, or from a bill, log or benchmark we produced ourselves. Hands-on articles describe systems we really run. No invented test drives.
AI Costs & Pricing
All →Choosing an LLM API Model by Cost Tier: Small, Mid, Flagship, Frontier
A practical framework for picking an LLM API tier by cost: worked examples for a classifier, RAG chatbot and long-form writer, plus routing and escalation.
Gemini 3.8 Flash Price Doubles Jan 1, 2027: What It Means for Bills
Gemini 3.8 Flash goes from $0.75/$3.75 to $1.50/$7.50 per 1M tokens on Jan 1, 2027. Worked bill examples, alternatives, and what to do before January.
AI Automation, Built
All →Anthropic Compliance API Adds Unified Chats and Word Document Export
Learn how the Compliance API update expands chat endpoints to unified Claude sessions and adds Word document downloads for Claude Docs.
Claude Dynamic Workflows: Managed Agent Background Runs
Configure Anthropic Claude dynamic workflows in Managed Agents. Learn how background runs orchestrate parallel agents, stream events, and handle error states.
Automated Blog Pipeline with Gemini: Architecture of What We Built
How our Node.js pipeline turns a brief into an SEO-checked Korean blog draft: autocomplete keywords, competitor research, schema-bound writing, lint, images.
Self-Hosting & Infra
All →Cloudflare Email Routing via API: A Working hello@ Address in Minutes
Set up a forwarding hello@ address with the Cloudflare Email Routing API: destinations, enabling the zone, a literal rule, DNS checks, and error 2007.
Debian 13 unattended install: a preseeded netinst ISO for a Dell R730
Remaster the Debian 13 netinst ISO with xorriso so it installs unattended (partitions, user, SSH keys, packages), then test it in QEMU and write it to USB on macOS.
Swapping Ghost for a Static Astro Site Behind a Cloudflare Tunnel
How we replaced a Ghost blog with a static Astro site on the same tunnel without touching its config: a Docker network alias, 410s for old URLs, and two nginx traps.
Models & APIs
All →PDF and Document Input in Claude, GPT, and Gemini
Compare document handling in Claude, GPT, and Gemini APIs: request payloads, file limits, non-PDF handling, and token implications.
Gemini 3.7 Flash Deprecation: Migrating to Gemini 3.8 Flash
Understand the 2026-10-08 Gemini API deprecations, automatic routing changes, thinking level updates, and migration steps for Deep Research agents.
Sending Images to Claude, GPT and Gemini: Formats and Limits
Compare image input methods, request structures, payload limits, and model token costs across Anthropic Claude, OpenAI GPT, and Google Gemini.