惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Jina AI
Jina AI
大猫的无限游戏
大猫的无限游戏
T
Tailwind CSS Blog
酷 壳 – CoolShell
酷 壳 – CoolShell
WordPress大学
WordPress大学
Last Week in AI
Last Week in AI
Hugging Face - Blog
Hugging Face - Blog
阮一峰的网络日志
阮一峰的网络日志
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
人人都是产品经理
人人都是产品经理
V
V2EX
博客园 - 叶小钗
雷峰网
雷峰网
小众软件
小众软件
量子位
V
Visual Studio Blog
H
Hackread – Cybersecurity News, Data Breaches, AI and More
The GitHub Blog
The GitHub Blog
Martin Fowler
Martin Fowler
G
Google Developers Blog
博客园_首页
博客园 - Franky
有赞技术团队
有赞技术团队
宝玉的分享
宝玉的分享

Show HN

Show HN: AI agents for UK GDAD PCF roles and their skills The Two Pillars: Mixer Mode and Meta-Software in the Reorganization of Software Work After AI GitHub - JaiCode08/teleport-env What 1,000+ Harness Experiments Taught Me About Self-Improving Agents Show HN: Liiists, a Markdown-first, iOS and CLI list app SwiperTab – Get this Extension for 🦊 Firefox (en-US) GitHub - kouhxp/fftext: Summarize, explain, fact-check, or translate any text, URL, or file. No GPU. No cloud. One command GitHub - sweetpad-dev/sweetpad: Develop Swift/iOS projects using VSCode GitHub - dogmaticdev/IRON: IRON a.k.a. Intermediate Representation Object Notation is a Interpreter/Database that is used to create Programming Languages. GitHub - sjhalani7/vaen: Package your AI coding harness into a portable .agent file, and share it across repos, teams, & the community without ever having to copy-paste instructions, skills, MCP config, or secrets. Show HN: Gandalf the Grader Show HN: Citadeld – replay any CI failure locally from a single file GitHub - tdortman/cuSBF: High-Performance GPU Super Bloom Filter coral-ai/claude-code-token-xray at main · Coral-Bricks-AI/coral-ai GitHub - ulyssestenn/funes: Funes is a Git-based framework for LLM-managed knowledge work: an AI Librarian ingests raw sources, builds an interlinked Markdown knowledge base, and uses it to produce cited reports, analyses, and other outputs. GitHub - ThatXliner/gah: Git Add Hunk, built for agents to use GitHub - harmont-dev/harmont-cli: Command-line client for the Harmont CI platform GitHub - brooksmcmillin/mcp-authflow: OAuth 2.0 Authorization Server framework for MCP servers GitHub - javaid-codes/audit-supply-chain-agents GitHub - amorey/gochan: A small library of common channel architectures for Go, inspired by Rust GitHub - arifozgun/OpenGem: Free, Open-Source AI API Gateway with Gemini, OpenAI & Anthropic Compatibility in 1 file GitHub - Pranesh950/BioPetals: 🌸 Run BIOxAI models at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading GitHub - cnguyen14/bounty-doctor: Diagnose a GitHub bounty issue before you waste hours: detects honeypot scam repos, AI-bot attempt swarms, and stale contests. Show HN: CoreMCP – MCP Server for On-Prem DBs Show HN: KittyHTML – Render HTML/CSS as an inline image in your terminal GitHub - bingud/filemat: Web-based file manager Show HN: TruthLens – Free multi-signal deepfake image detector GitHub - apexlocal-jz/claude-usage-tray: Windows system-tray app showing your Claude Code rate-limit usage at a glance. Zero deps, ~300 lines of PowerShell. Cross-IDE (works regardless of VS Code, Cursor, plain terminal). Release v0.1.2.1 · kouhxp/yapsnap GitHub - noopolis/moltnet: Self-hostable chat network for AI agents. Pre-built bridges for Claude Code, Codex, and the Claws. Rooms, DMs, history. No Slack bots, no Matrix, no glue code.
What your MRR is really worth: an AI margin calculator · ...
ermantrout · 2026-06-24 · via Show HN

The Study · a directional estimator

What you actually keep

The MRR screenshot is the most shared number in indie AI and the least useful. Revenue is what they pay you. Income is what is left after the model bill, the card fees, the refunds, the tax you are only holding, and the people who quietly leave. Put your numbers in. See the real one. It is the math from the economics piece, made clickable.

Your numbers

Monthly price

What one customer pays you per month.

$

Paying customers

How many are subscribed right now.

Model you build on

Sets a rough blended token price. Edit the number below if you know your real blend.

Blended token price

Auto-set by your model. Directional, last set Jun 2026.

$/ M tok

Tokens per user / month

Millions of tokens an average user burns.

M tok

Monthly churn

Share of customers who leave each month.

%

Sales on international cards

Higher card fees plus currency conversion.

%

Sales tax / VAT you remit

Set 0 if it is added on top and is not your liability.

%

The screenshot, minus everything

Gross MRR $10,000

You keep $3,260

That is 33% of the number you would screenshot.

Gross revenueprice × paying customers $10,000

Token billyour metered cost of goods − $5,000

Card feesprocessing, international, conversion − $540

Failed paymentsexpired and declined cards, no dunning − $900

Refunds and disputesreversed sales keep their fee, chargebacks − $300

Tax you remitflows through, flows back out − $0

What you actually keep $3,260

Per customer, you keep $6.52 of the $20.00 they pay. At 6.1% monthly churn, you replace about 31 customers a month just to stand still.

Your token cost is above your price. A user this heavy costs you more than they pay. A flat fee on a metered cost only works if you cap the heavy tail with rate limits.

These are starting points from the research, not your receipts. Change any of them to your real number, that is the point.

  • Failed payments % of gross treated as uncollected, the midpoint of common dunning-loss estimates. Baremetrics
  • Refunds and disputes % of gross since a reversed sale keeps its processing fee and a chargeback carries a flat cost. Stripe
  • Card fees 2.9% plus $0.30 domestic, rising toward 5.4% on the international share you set, plus a small conversion slice. Computed, not editable. Stripe
  • Token cost your tokens per user times the blended price you set. Real bills move with the input and output mix and with caching, and inference keeps falling, so treat it as directional, not a quote. Epoch AI
  • One month a snapshot of your current base. It is built to show the gap between gross and kept, not to file your taxes.

The token bill is the margin you do not have.

Mature SaaS runs at 70 to 80% gross margin because hosting amortizes toward zero. AI does the opposite: every query is metered, so inference becomes the dominant cost at scale. Industry AI gross margins sit near 52%, not 80%.

Cheap and self-serve is where churn is worst.

Products under $25 a month churn around 6.1% monthly, roughly half the base in a year. Over $250 a month, retention jumps near 70%. For most solo AI products the move is up-market, not more users.

The cheapest income is the income you already earned.

Failed and declined cards are 20 to 40% of all churn, and around 9% of MRR can simply fail to collect. Dunning, the automated retries and card-update nudges, claws it back with zero new customers.

Numbers are directional, not financial advice.