Seven years of GitHub history from one VP of engineering: AI didn't shrink his workload, it grew his appetite for work

Seven years of one engineer's GitHub history give the AI-productivity debate what its surveys never had: a measured answer. AI did not hand Orestis Ioannou his time back; it handed him more work to want. By October 27, 2026 he had logged 2,102 GitHub contributions, 88.7 percent of the 2,371 he recorded over the same stretch of 2020, when writing code was his whole job 1.

The dataset is his own pull, not a platform study. GitHub's contribution count mixes commits, pull requests, reviews and issues, private work included, and Ioannou calls it "a rough signal of activity, not a measure of value" 1. He joined Bold.org in January 2020 as a Full Stack Engineer, became Principal Engineer in August 2022 and Vice President of Technology in August 2025 1. On paper his graph should have kept sagging as he climbed. Instead it snapped back to full-stack pace; the essay explaining why is titled "AI Promised Less Work. I Have Never Worked More." and has practitioners arguing it out on Hacker News 1; 2.

What seven years of Ioannou's GitHub account show across three roles

The essay prints raw counts; the timeline below keys them to his roles 1:

• 2020, Full Stack Engineer: 2,711 contributions, still his busiest full year 1
• 2021: 1,839. 2022: 1,498. 2023: 1,189, his lowest full year 1
• August 2022 to July 2025, Principal Engineer: 114 contributions a month on average, against 173 a month as a full-stack engineer 1
• August to December 2025, first months as VP: 111 a month, with December 2025, at 50, the quietest month since he joined Bold.org 1
• 2026 through October 27: 2,102, already more than any full year from 2021 to 2025, with a record 427 in August 1

The essay prints the counts but not the divisions, so here they are. Across the 300 days through October 27, those 2,102 contributions run at about 213 a month, nearly double his Principal Engineer average of 114, in a job that is now leading engineers rather than shipping features 1. August's 427 sits 34 percent above the old monthly record of 319, which he set in April 2020 when coding was his full-time assignment 1. He has had 18 weekdays with zero recorded contributions (recorded GitHub activity, not days off) out of the 214 weekdays in that window, a pace that lands around 22 for a full year, against 40 to 46 in each full year from 2022 to 2025 1. The median weekday moved from 3 or 4 contributions to 7, near the 8 of 2020, and he already has 16 days at 24 or more contributions, versus exactly one such day per year from 2022 to 2025 1.

Starting got cheap, so he said yes to more

Ioannou's explanation is not that agents sped up his old workload. It is that the cost of starting something collapsed and his list of things worth starting grew to fill the space: bugs he would have left alone, pattern cleanups, projects that two years ago were not worth beginning 1. Almost all of it is private work; public contributions, near zero from 2021 to 2025, are 191 this year, under a tenth of his total 1.

The year came in bursts, not a climb 1:

• January and February: 277 and 233 contributions, when he began working seriously with agents and an observability experiment grew into Bold.org's internal company agent 1
• March to June: a wind-down from 203 to 77, with an April relapse of 187 that included a 29-contribution day and his biggest public month, 68 1
• August: a zero week after summer break, then every weekday from August 10 to 27 between 14 and 43 contributions, ending at the record 427, with September holding at 301 1

He counts three of these cycles since January, each ending when review queues and context switching catch up and he stops wanting to start another agent 1. That is the planning problem for anyone who manages builders: an agent that saves a person three hours does not bank three hours, it lowers the price of the next yes. "Time saved" measures the wrong thing; what moved is how much work feels worth taking on.

The n=1 limits, stated as plainly as the finding

This is one senior engineer at one company. Ioannou concedes the caveats himself: he uses agents for all of his engineering work and likes what they let him ship, which is exactly the motivated-reasoning shape a skeptic would look for; his 2019 baseline is incomplete because his previous employer used self-hosted GitLab; the counts are mostly private, so nobody outside can re-run them; and the essay itself was drafted with AI help, which he discloses while calling the experience and opinions his own 1.

The Hacker News thread splits where the essay does 2. llm_nerd calls the premise "a bit of a strawman" and asks who pitched less work in the first place 2. nkrisc argues there were only ever two outcomes: more work for the same pay, or layoffs with the work transferred to whoever stays 2. oxyg3n reads the chart as acceleration with no way off the pedal 2. arend321 answers from the other side of the fence, having cleared CRM tasks, accounting and a search-infrastructure upgrade by voice during a barefoot beach walk: "I may not work less, but certainly different and better" 2.

Almost all of that record is private, so the thread is arguing around a calendar it cannot open. That is both the essay's honesty and its ceiling. What the ledger does show is a Vice President of Technology running at nine-tenths of his own 2020 coding pace because his yes got cheap. His closing line is four words: "And I am tired." 1