GitHub COO Kyle Daigle posted the numbers this week: 1 billion commits in all of 2025, now running at 275 million per week, on pace for 14 billion this year. GitHub Actions compute jumped from 500 million minutes per week in 2023 to 2.1 billion minutes this week alone. Software demand does not saturate. Not every category of work can say the same.
Tomasz Tunguz maps this with a 2x2 matrix across two axes: whether demand is finite or infinite, and whether AI can close the verification loop without human judgment. The four quadrants produce different economic outcomes. Closed Loop plus Infinite Demand produces economic engines, software engineering being the clearest example. Closed Loop plus Finite Demand produces efficiency plays, think bookkeeping or tax filing, work that gets faster but does not multiply. Open Loop plus Infinite Demand produces creative amplifiers, content and marketing, where AI generates but humans still judge. Open Loop plus Finite Demand produces utility tools, 10-K prep, contract review, claims processing, one-and-done work that AI accelerates but cannot expand.
The matrix is worth reading in full because the real argument is about trajectory, not just current state. Some open-loop problems will close as AI verification improves, shifting entire categories from amplifier to engine. Tunguz also places venture capital explicitly in the finite-demand, open-loop quadrant, a pointed claim that the piece does not fully defend. That tension alone is reason to read the original.
[READ ORIGINAL →]