6/6
What’s next?

* Train newcomers to use AI responsibly.
* Automate reviews to reduce expert burden.
* Reward experts for mentorship, not just fixes.

Full paper: dx.doi.org/10.2139/ssrn...
Where do you see this happening in your field? #AI #Workplace #Innovation

5/6
Why it matters:

* Experts are a bottleneck. If they’re stuck fixing, innovation slows.
* Quality risks: AI can introduce errors that experts must catch.
* Burnout: Experts may leave if their role becomes only reviewing.

4/6
Beyond coding: This risk applies to any role where AI "helps" newcomers:

* Customer support: AI drafts replies → seniors edit.
* Design: AI generates layouts → experts refine.
* Legal/writing: AI drafts docs → seniors verify.

More output ≠ better quality.

3/6
The pattern:

* AI lowers the barrier → newcomers contribute more.
* But their work needs more fixes, overwhelming experts.
* Result: Experts innovate less, review more.

This dynamic could reshape any field where AI assists workers.

2/6
How we show this:

* Analyzed 2,755 open-source projects and 1,699 contributors before/after Copilot’s launch.
* Used difference-in-differences to isolate AI’s impact.
* Found: Experts shifted from coding to reviewing, while AI-assisted work needed more rework.

Earlier this year, we used less granular data as a basis to model the resulting dynamics. Such as periods of overordering of GPUs.

Will be interesting to see how this will develop.

2/2

papers.ssrn.com/sol3/papers....

Blijkbaar hebben ze meer vertrouwen in de Zweedse wetgeving en bedrijfsvoering

Op zich wel handig dat ie de lijn stopzet nu er ook enorme tarieven op datzelfde staal worden gezet 🤷

Welcome! I did the same move a few weeks ago after a similar hibernation period

Trump's deportations of immigrants working on dairy farms become a significant contribution to fighting climate change #emissions #secondordereffect #unintendedconsequences