Resources

Insightful resources for the Consumer Dividends professional, free of charge to keep forever.

We publish resources for people carrying the work of a whole team on their own. Everything on this page costs nothing, asks for no email, and creates no account. Resources like these are an aspect of our devotion to our Core Four principles.

Reading and writing Open datasets

Reading and writing

Free newsletter

Dividends

Reading the filing to work out who gets paid.

Andrey writes Dividends, working through S-1s and market events with the source document open and the figures quoted, so a reader can check the claim instead of taking it. Three pieces are up: the Jersey Mike's listing at up to $12 billion after Blackstone, SK Hynix going to the Nasdaq at a trillion-dollar valuation on the back of Nvidia memory demand, and what the SpaceX filing shows about governance concentration and AI spending.

S-1 breakdowns drawn from the filing itself, because the press release and the prospectus rarely say the same thing
Who collects in a deal, meaning the sponsor, the insiders, and whoever is holding shares after the lockup expires
Founder notes from building a product studio alone, including the parts that went wrong and what they cost
Finance and news commentary, with the occasional satire

Freeread it without subscribing

Read Dividends See the archive

Free guide

The AI Slop Field Guide

How to spot language-model writing, and how to stop producing it.

More than half of what you read online started as model output. This teaches you to recognize it by four mechanics: word choice, sentence structure, paragraph shape, and argumentation. If someone is asking for your attention or your money on the strength of their thinking, you should be able to tell whether the thinking is theirs.

Four categories of word choice that show up disproportionately in generated text
Six sentence patterns models fall into, and why the process produces them
Paragraph and argument tells, including the ones that survive a careful edit
A sixty-second scan for vetting any piece of writing, including your own drafts
A prompt you can paste in to stop a model from producing it

Freeno email, no signup

Read the guide Download the PDF

Open datasets

Two synthetic conversation datasets, published open on Hugging Face for anyone training or evaluating a model in these domains. Every record is generated, so none of it is real people or real matters, and every record passed a 1.0 quality gate before publication. Released under CC BY-NC 4.0.

Open dataset

HR Conflict Dataset

500 labeled workplace-dispute conversations.

Synthetic dialogue covering workplace conflict and dispute scenarios, with the distribution anchored to published EEOC charge data and Bureau of Labor Statistics wage figures rather than invented from nothing. Built for anyone training or evaluating a model on employment-relations language.

500 records, labeled, with a dataset viewer
Distribution anchored to EEOC charge patterns
Wage figures anchored to BLS published data
Fully synthetic, no real matters and no real people

FreeCC BY-NC 4.0

View on Hugging Face

Open dataset

Customer Churn Dataset

500 labeled retention and cancellation conversations.

Synthetic dialogue covering the conversations that happen when a customer is deciding whether to stay, anchored to published Recurly and ProfitWell retention benchmarks. Built for anyone training or evaluating a model on retention and customer-success language.

500 records, labeled, with a dataset viewer
Distribution anchored to published retention benchmarks
Fully synthetic, no real customers and no real accounts

FreeCC BY-NC 4.0

View on Hugging Face

Everything we have published is on the Consumer Dividends Hugging Face profile.

We build the same way we publish.

These are free because the people we build for rarely have a budget for the thing that would help them most. What we sell holds to the same standard this page does: take a position, show the number, and let someone check the work.

See Holdings