All articles
The Calcaas Blog

All articles

Page 2 of 11

Log-Scale Pricing Charts Are Quietly Wrecking Your Model Choice
Founder Guides
Sep 5, 20266 min read

Log-Scale Pricing Charts Are Quietly Wrecking Your Model Choice

The intelligence-versus-cost chart everyone quotes uses a logarithmic price axis, which compresses a 50x difference into a couple of centimetres and makes wildly different bills look like neighbours.

A Frontier Model Launch Is a Repricing Event: A Six-Step Checklist for Founders
Pricing Strategy
Sep 5, 20266 min read

A Frontier Model Launch Is a Repricing Event: A Six-Step Checklist for Founders

When a provider ships a new flagship at a new rate, your cost of goods sold changes before your pricing page does, so treat launch day as a margin review rather than a feature announcement.

Cost Per Token vs Cost Per Task: Why LLM Pricing Comparisons Keep Misleading Founders
LLM Economics
Sep 5, 20266 min read

Cost Per Token vs Cost Per Task: Why LLM Pricing Comparisons Keep Misleading Founders

A model priced 2.5x higher per token can still produce a smaller bill, because what you actually pay for is tokens consumed per task, not the rate on the pricing page.

Meta Just Put a Price on Your Prompts, and It Is Not 95%
Pricing Strategy
Sep 4, 20266 min read

Meta Just Put a Price on Your Prompts, and It Is Not 95%

Meta's Muse Spark contributor tier cuts input tokens by 92% and output tokens by 95.3%, so the discount you actually capture depends entirely on your input to output ratio.

Compute Capacity Belongs in Your AI Cost Model, Not Just Your News Feed
LLM Economics
Sep 4, 20266 min read

Compute Capacity Belongs in Your AI Cost Model, Not Just Your News Feed

When a provider adds capacity and raises your rate limits, your effective cost per user can fall even though the list price per token has not moved at all.

A Cheaper Model Does Not Lower Your AI Bill: What Fable 5.1 Actually Changes
LLM Economics
Sep 2, 20266 min read

A Cheaper Model Does Not Lower Your AI Bill: What Fable 5.1 Actually Changes

Anthropic's Fable 5.1 is built to reduce token cost, but your monthly AI bill is set by cost per completed task, not by the number on the model card.

OpenAI Is Selling Ads on ChatGPT's Free Tier: What It Means for Your AI Pricing Model
Pricing Strategy
Aug 28, 20264 min read

OpenAI Is Selling Ads on ChatGPT's Free Tier: What It Means for Your AI Pricing Model

OpenAI has started showing ads to ChatGPT's Free and Go users in India, turning ad revenue into a third monetization lever alongside subscriptions and API usage, a move most AI builders can't copy but should still learn from.

OpenAI's Jalapeño Chip and What Custom Inference Silicon Means for LLM Pricing
LLM Economics
Aug 26, 20264 min read

OpenAI's Jalapeño Chip and What Custom Inference Silicon Means for LLM Pricing

OpenAI's custom Jalapeño inference chip beat the current state-of-the-art Nvidia system on tokens served per user and throughput per kilowatt, a result that points toward cheaper inference and more pressure on provider pricing.

A 4-bit Model Just Beat Its Full-Precision Original: What That Means for Your Inference Bill
LLM Economics
Aug 26, 20264 min read

A 4-bit Model Just Beat Its Full-Precision Original: What That Means for Your Inference Bill

Multiverse Computing's Quantization-Aware Healing technique made a compressed, 4-bit LLM outperform its own full-precision checkpoint on 7 of 9 benchmarks, while using a fraction of the memory and compute per token.

I Modeled What $20,000 a Month on Devin Actually Buys a Solo Founder
Founder Guides
Aug 25, 20264 min read

I Modeled What $20,000 a Month on Devin Actually Buys a Solo Founder

A solo founder's public breakdown of spending $20,000 in a month on the Devin coding agent is a useful stress test for how any founder should think about AI-agent ROI, not just whether the number sounds high.

Good Enough and 100x Cheaper Is Beating Best-in-Class: The Economics of AI Simulation
LLM Economics
Aug 25, 20264 min read

Good Enough and 100x Cheaper Is Beating Best-in-Class: The Economics of AI Simulation

A widely cited breakdown claims simulation-based methods trade a 10% quality hit for 100x lower cost and 10,000x faster iteration, and that math should change how you think about "best" model selection.

GPT-5.6 in Kiro: What "Better Price-Performance" Actually Means for Your Token Bill
LLM Economics
Aug 25, 20264 min read

GPT-5.6 in Kiro: What "Better Price-Performance" Actually Means for Your Token Bill

OpenAI says GPT-5.6 improves price-performance for developers inside Kiro, but the number that decides your margin is cost per finished task, not price per token.

The Margin Memo

Pricing math, in your inbox.

One short note a week on AI pricing, token economics, and margin. No spam, unsubscribe anytime.