About the author
Jason Gong here. Saturday Robot ships each week with what the most serious AI builders are actually doing. Before working on this and doing fractional growth work, I ran growth at Kite, an AI coding assistant from before ChatGPT (7M users, acquired by Affirm), founded Firezone (YC W22), and led GTM at GrowthX, where we built AI growth machines for companies like Lovable, Webflow, and Surge AI ($12M in 9 months).
This February, my tokens cost more than a headcount for the first time. About $5,000 in a single month.1
It sounds like a lot until you look at what the month contained. I spent more time in Claude Code and Codex than in any of the tools that used to define my job. I wrote strategy in markdown. I reviewed copy in pull requests. Agents sent emails, wrote blog posts, created models, and set strategy. The people around me were doing work that nothing in their job titles would have predicted.
The work is changing faster than job titles. That is the piece that has been sitting with me, and I think a lot of people recognize the feeling without being sure what to do with it.
Coding's head start: code was already canonical, public, artifact-leaving
Almost eight years ago I was head of growth at Kite, an AI coding assistant that existed long before ChatGPT.2 The models back then could autocomplete and hint but not really reason with you. We charged for tokens before most people knew what a token was.
What the Kite era made obvious in hindsight is that software absorbed AI fast because the substrate was already right. Engineers had spent decades putting their work into text. Code was canonical. Documentation mattered. Pull requests were public. Mistakes left a record. Once the models got good enough to reason, those habits started doing real work.
Marketing was the opposite. Strategy lived in decks that went stale. Positioning was a paragraph somebody remembered from an offsite. Playbooks were PDFs nobody opened. The work left almost no durable artifact a model could read.
The same shift is now happening in other functions
Last year at GrowthX I built a go-to-market team where three of four people had never opened a terminal. By the end of the year they were working out of a shared repo. Copy got reviewed in diffs. Playbooks lived alongside the deliverables they produced. One junior hire who had been writing in Google Docs six months earlier shipped a full positioning rewrite as a PR with suggested commits from three different reviewers. He kept asking me if that was normal.
The compounding showed up fast. A positioning doc sharpened a landing page. The landing page improved an outreach sequence. The sequence surfaced a new objection, which updated the positioning. The repo stopped feeling like storage and started feeling like infrastructure. Revenue across those engagements grew roughly 7x over the year.3
When Block cut roughly 4,000 people this February, Jack Dorsey wrote that intelligence tools paired with smaller and flatter teams were changing what it means to run a company.4 Every layoff is not downstream of AI. But the market has the incentive to tell that story, and stories like that harden into strategy before the technology catches up.
Why write about this?
The best thinking I have found on any of this is from practitioners who write as they go. Paul Graham on how writing is thinking.5 Simon Willison shipping a tool and posting about it the same week.6 Ben Thompson naming the structural force underneath an earnings call.7 Most AI coverage is a 101 explainer or a 2030 prediction. The middle, where you decide what to try on Monday, is thin.
I ran monthly workshops with thousands of registrants this past year to teach people how to use AI in content marketing. The people who ramped fastest all had some version of a shared system underneath them. The people still stuck were running the same prompt-per-task loop they had been running six months earlier. Nobody had shown them what the setup should look like once it was working.
The promise: have something to try Monday
Here is my promise with Saturday Robot. Every post will be about the work itself. How a thing got done, what made it hard, what broke, what I am still figuring out. Mostly my own field reports, sometimes people whose setups taught me something. If I do not have something substantial to say, the piece will be short. If I do, I will go deep. Either way, by the end you should have one concrete thing you can try in your own work that week.
This post is no exception. Alongside it I am launching a starter kit for a knowledge base you can run agents on top of.8 Templates and examples you can copy and have running in an afternoon. By Monday you can have your first agent reading your own work.
Saturday is when you build what you want. The robot is the thing you are building. Something that did not exist on Friday and works by Sunday.
If any of the feeling I started with sounds like your last year, I think you are in the right place.
A note: the starter kit goes live Monday, May 11. The site is still rough in a few spots. I am fixing things as I go.
Field notes on growth from the best operators
Companies with unfair distribution win. I help unpack what top operators are learning and investing in to help you think through growth.