Craig CossairtIndependent by nature.
← Back to Craig
Sometimes, the output is words.

Thinking
in public.

Twice a week, I write for Bloom about relationships, technology, and the choices behind them. These are the latest.

Dating and relationship advice

A situationship has the affection. The research says it does not have the commitment.

An exploratory 2024 study defined situationships as affection, sex, and time together without a label or much commitment. A 2006 longitudinal paper found that swings in how committed you think your partner is predict breakup beyond the average level. The slang is new. The measurement is not.

Intentional dating

77 percent of singles say dating feels like a marketplace. Two-thirds have already left the apps.

Harris Poll's Modern Dating Survey found 77 percent of singles say dating feels more like a marketplace than a romance story, and two-thirds of singles are not currently using dating apps. People are not leaving romance. They are leaving a search format that feels like inventory.

The lifecycle journey

What your partner does with your good news predicts more than how they handle your bad news

Relationship advice prepares you for the argument. In one study of 79 dating couples, how partners responded to each other's good news was more closely related to relationship well-being two months later, and to breaking up, than how they responded to bad news. Three of the four ways to get it wrong feel reasonable from the inside.

Dating and relationship advice

People who believe in 'the one' are more likely to ghost. The research suggests why.

The first research to ask what predicts ghosting went looking at beliefs rather than character. People who think partners are either compatible or they are not felt better about ghosting, intended to do it more, and reported having done it more. Cruelty and cowardice were never measured, so nothing here rules them out.

Aligned monetization

Bumble deleted the rule it was built on. The rules that cost money go first.

On August 11 the app that spent a decade on one idea stopped requiring women to message first. The company gave a user-research reason and the reason is probably true. It is also true that the rule suppressed the exact number a dating app is paid on, and that this is the kind of principle that never survives a bad year.

Intentional dating

Three in ten singles used LinkedIn for dating last year. The reason is not romance.

Singles are running background checks on each other through a professional network, and in one survey of US employees, 48 percent said they trust a LinkedIn profile more than a dating profile. What they are reaching for is not a better feed. It is a name attached to a history that somebody else can check.

Utah launch cluster

What 'best dating app in Utah' actually has to mean

Ask for the best dating app in Utah and you get a national ranking with a state name taped on. Then, if the answer knows this market, Mutual. What the question actually requires, and why a Salt Lake City list that ignores the Wasatch Front is answering a different place.

Dating and relationship advice

Love Is Blind works. That is exactly the problem.

The pods are the part with research behind them. Escalating mutual disclosure reliably produces closeness fast, and that is the one thing the show does well. The failure is what it asks people to do with that feeling, on a deadline that makes the actual predictors of a lasting relationship impossible to observe.

Relationship growth science

Americans did not stop wanting partners. They stopped believing the search works.

The romance recession is the story of the moment. Then you check the one number that would prove it and find a twenty-year trend quietly reversing. What actually declined was not partnership or desire. It was faith in the method.

Dating and relationship advice

Your attachment style is not a personality type

The internet turned attachment into four boxes and a quiz you take once. Researchers do not measure it that way, and a study of more than 4,000 people found it moves. What the science actually says, and the finding that tells you when it moves.

Intentional dating

Machine learning cannot predict chemistry. The same method predicts relationship quality.

Researchers collected more than 100 traits and preferences from every participant before anyone met, then asked a machine learning model who would want whom. On that question it returned nothing. The same researcher, using the same method inside real relationships, found the opposite.

The lifecycle journey

Technoference: what the research says a phone does to a relationship

Researchers have a name for the phone coming out mid-sentence, and a decade of studies on what it does to couples. One 2023 result should change how you handle it, because it points somewhere other than screen time.

Originally published on Bloom. Dates are the original publication dates.

Pre-launch
Relationships

Bloom

A dating app built for the relationship that comes after the match.

Bloom’s public website introducing its dating and relationship experience

The question

Finding someone is only the beginning. Dating products tend to focus on the search, while the work of building a relationship happens somewhere else.

What I built

I’m building Bloom around the whole relationship journey. Dating Mode helps people meet with intention. Relationship Mode gives couples a shared space to plan dates and keep growing together.

Where it stands

Core matching tools, including preference filters and seeing who liked you, are free. AI-assisted features are labeled and optional. Bloom is preparing for launch in Utah; the public waitlist is open.

Solo founder, with AI as a working partner.

Visit Bloom
Live product
Family & imagination

Cozy Fables

Personalized bedtime stories with your child at the center, made to read together.

Cozy Fables public website showing its personalized bedtime story experience

The question

Storytime is a chance to make something together. A familiar name, a favorite world, and a little choice can turn a bedtime story into a shared adventure.

What I built

I created Cozy Fables, an independently run family product that generates personalized bedtime stories. Parents choose the characters, world, style, and reading level; each story is made to be read together.

Where it stands

The browser-based product offers a free first story and a Family plan. Read-aloud prompts, branching adventures, saved stories, and printable PDFs give each family a few ways to make storytime their own.

Created and independently run by Craig Cossairt.

Visit Cozy Fables
Open source
Tools for builders

Trellis

The working structure behind an AI-assisted project, shared with other builders.

Trellis public repository and its guide to building with AI coding agents

The question

Building with AI gets complicated as a project grows. Decisions disappear, conventions drift, and the same mistakes can return in the next session.

What I built

Trellis turns the practices I use into a starter that other founders can adopt: shared agent instructions, a setup checklist, development methods, decision logs, and guardrails.

Where it stands

The core is plain Markdown and works across coding tools. Optional hooks and a local knowledge base add structure as a project needs it. The template is MIT licensed.

A public template extracted from my daily development setup.

Explore the source
Open source
Everyday life

Hestia

Meal planning, recipes, and a grocery list that knows what’s already in the kitchen.

Hestia’s public meal planning project and documentation

The question

A meal plan has to fit the household, the pantry, and the reality of getting dinner on the table. Those pieces are often scattered across separate tools.

What I built

Hestia brings nutrition targets, an AI coach, inventory-aware groceries, and recipe-to-cooking flows into one progressive web app.

Where it stands

Named for the Greek goddess of the hearth, Hestia is an open-source project built with Next.js and Supabase. The repository includes instructions for running your own instance.

An independent side project, shared as open source.

Explore Hestia
Open source
A useful little tool

Longshot

Full-page screenshots that stay on your machine.

An authentic screenshot of Longshot’s annotation and export editor

The question

Full-page captures can repeat sticky headers and miss content inside scrolling panels. Getting one clean image shouldn’t become a project of its own.

What I built

Longshot shares a capture engine between a Chrome/Brave extension and a headless CLI. It handles sticky headers and inner scrollers, then offers cropping, annotation, and export.

Where it stands

Screenshots are processed locally. The extension currently installs by sideloading from the repository; a Chrome Web Store listing is not published. The CLI supports automated capture workflows.

An independent tool with an MIT-licensed source repository.

Get Longshot
Hackathon winner
Public-interest technology

Privacy Workbench

A working demonstration of AI-assisted government data governance.

The Privacy Workbench demonstration interface

The question

Public agencies need to explain what information they collect, why they retain it, and who can access it. Consistency across offices is a difficult, detailed task.

What I built

I built the Workbench frontend and demonstration; Valerie Adams developed the agent orchestration engine and architecture. The interface brings provenance, review queues, and sensitive decisions into the officer’s workflow.

Where it stands

Our team won the Utah AI Data Protection & Governance challenge at the 2026 HITLAB/UVU event. The public demo replays curated data. Some scenarios are illustrative; it is a hackathon demonstration, not a deployed government service.

Craig Cossairt: frontend and demo. Valerie Adams: agent orchestration engine and architecture.

Open the demo
Two bounty builds
Tools for founders

Startup State

Two connected explorations of a simpler question: what should a Utah founder do next?

Startup State public demonstration of its Utah founder resource platform

The question

A large catalog of programs is only useful if a founder can find the right next step. The same challenge appears when searching for government opportunities that fit a particular business.

What I built

Across two AI Builder Day bounty submissions, I built a founder platform with a playbook, resource discovery, startup map, and an AI navigator, then a second experience that ranks federal and Utah opportunities by fit.

Where it stands

The two builds are presented together here. Startup State is a working prototype; Opportunity Finder explains possible program fit rather than promising eligibility. Neither is presented as an official government service.

Built by Craig Cossairt with AI coding tools. The first build uses resource and startup data supplied for the GOED challenge.

Explore Startup State
Bounty prototype
Healthcare coordination

BetterRX

A shared view of hospice equipment orders, from admission through pickup.

Public BetterRX prototype documentation and challenge notes

The question

Hospice teams and equipment vendors need to coordinate deliveries, changes, and pickups. A missed handoff can leave people chasing the same order in different places.

What I built

For the BetterRX DME Ordering and Visibility bounty, I built a phone-first equipment board with role-specific views, order tracking, and an oversight surface for the care team.

Where it stands

Created for AI Builder Day Part 2 in August 2026. The demo uses sample orders and includes scenarios for missed discharges, delayed pickups, and equipment costs. It is a bounty submission, not a deployed clinical system; no prize or contract award is claimed.

Built by Craig Cossairt using the BetterRX challenge brief and supplied sample scenarios.

Explore the prototype