I tried Computer History
Isaac's week: "too many tabs, somehow productive"
A note from me
Computer History is a fun thing to try. It gave me a cool way to look back and see just how productive my week was.
Computer History's response
Isaac spent the week operating less like one developer and more like a small, unusually determined engineering department.
Main quest: ship software, interrogate everything
A healthy chunk of the week went into building and improving software across several projects:
- Diagnosed a slow web app using browser traces and runtime timing
- Tested analytics capture with a physical Android device
- Built a substantial employee administration and data-import workflow
- Reviewed, repaired, and merged a small mountain of pull requests
- Improved his automated QA infrastructure
- Audited old code before removing it, narrowly avoiding deleting a feature that was quietly still live
- Investigated sharing web functionality with a desktop app, because maintaining two versions of the same page is how software slowly becomes haunted
QA became its own product
At some point, improving the QA process became a full engineering project of its own.
Isaac worked on:
- Persistent QA agents
- Automated follow-ups when pull requests receive feedback
- Testing for deployed sites
- Better access controls and failure reports
- Real-device validation
- Shareable QA artifacts
One desktop build produced the classic engineering result: packaging passed, updates passed, and the button people actually needed returned an HTTP 502.
Green enough to reveal the next problem. Not green enough to lie about it.
Surprise expansion pack: Codex Ambassador
Isaac also joined the Codex Ambassador program.
Naturally, he responded by immediately starting to plan a free, hands-on Codex workshop in Utah Valley. He researched venues, contacted community partners, worked through scheduling, and found a promising path for a space that can hold around 30 to 40 people.
You give Isaac one new role and suddenly he's doing event logistics between pull-request reviews.
Meetings, coordination, and assorted boss battles
Elsewhere, Isaac:
- Worked around a GitHub outage
- Coordinated across enough communication apps to qualify as a distributed system
- Started consolidating multiple calendars
- Set up some new productivity tools
- Did homelab maintenance
- Shut down a couple of unused services
- Played about 40 minutes of a game, providing the week with a rare period where the only observed architecture was WASD
This week's unofficial awards
- Most persistent antagonist: authentication
- Most improved supporting character: automated QA
- Best real-world proof: data arriving from an actual phone
- Largest spreadsheet encounter: hundreds of employees and nearly 100 locations
- Most suspicious sentence: "Why does this page still take ten seconds?"
- Best side quest: planning a free Codex workshop
- Final boss: context switching
So, how much work was it?
My estimate is roughly four pre-AI developers' worth of work, give or take one.
That comparison is fuzzy. AI makes implementation and investigation much faster, but it doesn't eliminate judgment, testing, or follow-through. Counting every agent task as independent human labor would badly inflate the number.
The more honest comparison is that one developer can now operate more like a small engineering team.
Producing code is getting cheaper. Knowing what should exist, tracing the real system, testing the result, catching bad assumptions, and recognizing when "done" is lying to you still require a human.
So perhaps the best description isn't "four developers."
It's one Isaac operating a mildly alarming control room.