Doubling a software team rarely doubles its output. Small teams punch above their size, coordination cost eats the gains as they grow, and a team of 100 that grew there is not the same as one of 100 that shrank from 150: output depends on the path, not just the size today. This tool traces that curve and lets you test how AI changes it: how far each person’s effective output rises, how the best team size shifts, and what a cut actually costs. It separates three kinds of cut (a routine layoff, a real handoff of work to AI, and a budget cut dressed up as AI), because each hits coordination, morale, and the cost of checking AI differently. Move any control and the chart stays in view; each one shows its equation, its sources, and how strong that evidence is.
The loop: the org climbs the growth curve as it hires; after a layoff at its peak size it returns down a lower contraction curve, never retracing its path. The shaded gap between the two curves is the hysteresis: output a given headcount keeps on the way up but loses on the way down. The readout below gives the effective-labor drop for the current cut.
{{ readoutsEl }}This measures the software org's own output (features and code shipped), not the output of the whole company the org sits inside. The two are very different numbers and were always disconnected: task-level dev evidence supports double-digit AI gains on the org's own work, while firm-wide productivity nets out to low single digits once diluted across non-engineering functions, integration lag, and the full cost of AI. Read every figure here as the engineering org's output; a company-level number would be substantially smaller.
Each mechanism is a toggle plus its strength. Turn one off to see it drop out of the curve.
The model's weakest joint. Each combines separate published literatures into a coupling; the join is the inference, not any single reported finding. See the methodology section for how each is assembled. Every one has a zero/identity value that collapses it.
The staffing question: what does it cost to ship the same output, staffed each way? Hold output fixed and compare the annual bill for an all-senior team, your mix, and an all-junior + AI team, each staffed to the size that delivers it. This is where junior vs senior comp finally bites: cheap juniors need more heads and capture less of AI’s lift, so the “just hire juniors + AI” saving often reverses. (Reaching a size by cutting instead adds the hysteresis penalty, the contraction curve on the chart.)
{{ econReadoutEl }}Overlay a second set of curves without losing the base loop.
Draw one law's signature shape on the chart; the loop stays in place. Toggle any number.
Return every control, toggle and layoff type to its default value. Individual tabs each have their own group reset.
If you use or reference this tool, please cite it. Swap the URL for wherever you host it (your site, or a GitHub Pages repo).
@misc{shkolnikov2026softwareteams, author = {Shkolnikov, Yakov Pyotr}, title = {Impact of AI and Layoffs on the Efficiency of Software Teams}, year = {2026}, howpublished = {\url{https://yakovshkolnikov.com/employment_july_2026.html}}, note = {Interactive model}
}