Write
Working Overtime
The lived experience of AI at work: ambition, ease, voice, compulsion, and agency.
Katie Parrott / Every / October 2024 - now
I started by writing about AI. Then it changed the job I had to do.
Working Overtime began as my dispatch from inside a changing working life. It led into a larger role at Every: testing new models, writing practical guides, building tools, and helping readers figure out how to use AI deeply without handing over their judgment.
The through-line
Every works at the frontier. I translate that frontier into agency.
I put the newest capabilities into an actual writer's working life, notice the relief and the failure modes, and turn the result into something readers can use. Sometimes that is an essay. Sometimes it is a guide, a benchmark, an app, or an argument with an AI pope.
Test & teach
Vibe Check + Guides
What the models can do now, and the workflows that make those capabilities usable.
Build
Tools + experiments
Tastemaker, agents, and interactive projects that make the argument concrete.
Recent work across Every / Spring 2026
The role now moves between story, system, and test.
In one month I wrote about entry-level careers and agent-managed workweeks, helped readers set up Codex for knowledge work, and tested a new frontier model. Just before that, I showed my full writing-with-AI process, including the parts I still refuse to delegate.
Vibe Check: Opus 4.8—Anthropic Should've Rounded Up to 5
Testing the frontier model on engineering, writing, and everyday knowledge work.
Codex for Knowledge Work
A setup and seven-day plan for turning an agent into support for research, planning, writing, and reporting.
How to Start a Career When AI Is Doing Your Entry-level Job
Four pieces of advice for gaining judgment when the old training ground is shifting.
I Let ChatGPT Manage My Workweek
An AI project manager with my OKRs, calendar, Notion, and Slack—and the parts it still cannot do for me.
Writing With AI Is Harder Than You Think
The receipts from my AI writing process: named reviewers, standards, pushback, and the judgment that keeps the words mine.
My Every arc / selected stops
I came in through a column. The job kept growing.
Working Overtime now runs to 25 essays. Along the way, writing about AI opened into building editorial systems, testing models, making products, and translating Every's frontier work into something readers can take into their own jobs.
-
01 Oct 2024 - Feb 2025
I entered Every through the human question.
Working Overtime started with career panic and a chat window. The early question was intimate and enormous at once: what happens to a knowledge worker when intelligence gets cheap?
-
02 Mar - Jul 2025
Writing about capability made me a builder.
I was reporting on AI's effect on jobs while using it to cross the boundary of my own. A literature major became the person building workflow tools for Every in public.
-
03 Aug - Nov 2025
My editorial taste became infrastructure.
The reporting and the making converged. I encoded Every's standards, trained writing partners on my voice, ran models in parallel, and wrote about the support that made work possible.
-
04 Jan - Apr 2026
The experiment became an argument about agency.
Building more with AI gave me better evidence about its cost. I wrote through compulsion, AI writing, and autopilot: the moment when a useful tool gets fluent enough to hide your own absence.
-
05 May 2026 - now
Now I work across Every's AI ecosystem.
The latest work reaches beyond the column: agent-managed workweeks, career advice for an automated era, Codex workflows, model tests, and cultural arguments about automation. The recurring question is still which decisions remain ours.
Where the work lives now
Every gave the experiments more than one form.
The first-person column remains home base. Around it is the practical layer of my work: guides people can follow, tests of the models they might use, and products or interactive projects that make an idea tangible.
Use agents
Codex for Knowledge Work
Set up an agent workspace for planning, research, writing, and reporting.
Read the guide →Keep taste
Tastemaker
A writing tool I built to make your preferences explicit before a model edits them away.
Explore Tastemaker →Test models
Vibe Check
Every's hands-on tests of whether a new model improves the work people do.
Read Vibe Check →See the stakes
The AI Pope, My Boss, and Me
I build the AI the Pope is warning about. My boss makes the optimistic case. I put all three of us in the argument.
Start the dialogue →Katie at Every
The tools got more capable. My job became helping people use that capability on purpose.
Working Overtime is where I narrate the change from inside it. The rest of my work at Every is where I test, build, and hand over what I learn.