Field Notes
Cost notes from building beforeyouship
Short, data-backed posts on LLM cost — what the pricing tables don’t tell you, modeled with real numbers. Every post links into the calculator so you can run it yourself.
2026-07-11 · Deep-dive
Chatbot LLM Cost: Stateless vs With History
Memory looks like a 2× tax — it's actually quadratic. Every turn re-sends the transcript, so long conversations carry the bill. Caching collapses it back to ~1.3×.
2026-07-03 · Deep-dive
LLM Model Routing: Cost Optimization Without the Quality Hit
Route easy requests to a cheap model and escalate the hard 20% — roughly half the bill, same quality floor. The catch: one number decides whether it pays off.
2026-07-02 · Deep-dive
The Cheaper LLM Isn't Always Cheaper
Your cost ranking is a snapshot at one retry rate. Because retries bill uncached tokens, the cheaper model can flip — the new sensitivity tripwire shows exactly when.
2026-06-27 · Deep-dive
The Real Cost of AI Agents (Multi-Step)
A single agent task fans out to 5–12 model calls — landing ~13× a chatbot. Step count and cascading retries are the two cost drivers nothing else has.
2026-06-25 · Deep-dive
RAG Pipeline Cost: Why It's 3–5× a Chatbot
RAG looks like one feature but it's 3–5 LLM calls per question — and with bigger prompts the bill lands ~6× a simple chatbot. Where the money goes, and how caching claws it back.
2026-06-18 · Deep-dive
Prompt Caching: How Much It Actually Saves
The discount is ~90% — but a write premium and your real hit rate decide whether caching cuts 60% off your input bill or nothing.
2026-06-11 · Guide
Query LLM Costs From Claude Code (Without Leaving the PR)
beforeyouship now runs a remote MCP server. Three tools put the full cost model — retries, caching, cascading agent failures — one tool call away from your editor.
2026-06-10 · Guide
How Much Does an LLM App Actually Cost?
Most teams underestimate by 2–3×. Here's the real math — retries, caching, and growth included — modeled across six models.