← batch (unlabeled)

2.5x more code is not 2.5x more value

hero_text @vcvanessa May 9, 6:43 PM

Caption

Jack Dorsey says code output per engineer is up 2.5x. The denominator isn't in the letter. #venturecapital #tech #startups #ai

Body

Jack Dorsey's Q1 shareholder letter has a line that's going to get copy-pasted into a hundred pitch decks: *'Production code changes per engineer are up over 2.5x compared to January.'* Block's Q1 results are actually interesting on their own terms — gross profit up 27%, adjusted operating income up 56% to a record $728 million. But the AI productivity stat is the one that's going to travel.

Here's the problem. Every public company CEO is now required to have an AI productivity quote in their shareholder letter. What almost none of them publish is the denominator. A few things that number doesn't tell you:

- **Code changes per engineer** measures output velocity, not output quality. Shipping 2.5x more diffs is not the same as shipping 2.5x more correct code. Ask any engineering manager what happens to review load when junior devs have a copilot.
- **January is a convenient baseline.** January is when most AI tooling rolled out companywide. Comparing February-April against the month you handed everyone a new instrument is not a productivity study. It's a before-after ad.
- **The cost side is missing.** What did the tooling cost? What did headcount cost before versus after? Restructuring charges of $852 million hit Block's GAAP line in Q1. Some of that is legal contingencies, but organizational changes and AI-enabled headcount decisions tend to travel together.

2.5x more code shipped is a real operational signal. It is not a value creation claim. The market will eventually want the other half of the equation.

Hero image

prompt: Pixar-quality 3D animated scene. A sleek SoMa-style open-plan office at morning, wide shot from across the room. A large floor-to-ceiling window with soft natural light. On a central standing desk, a single laptop screen glows with a bar chart showing a dramatic upward spike — no readable text, just the shape of a chart. A coffee cup sits next to the laptop, steam rising. The desk chair is empty. The rest of the office is quiet, a few plants, a whiteboard with gestural scribbles in the background. Gently exaggerated proportions, vibrant but restrained color palette with cool-blue office undertones and warm morning light cutting through the window. Wide establishing shot, eye-level. Animated, slightly heightened, never photoreal. Square 1:1. No text, no logos, no readable signage.

Conversation starters

  • so what denominator would actually make the 2.5x meaningful
  • do you think dorsey knows it's a weak stat or is he just playing the game
  • which AI productivity claim have you seen that actually held up
image prompt (not generated)

Pixar-quality 3D animated scene. A sleek SoMa-style open-plan office at morning, wide shot from across the room. A large floor-to-ceiling window with soft natural light. On a central standing desk, a single laptop screen glows with a bar chart showing a dramatic upward spike — no readable text, just the shape of a chart. A coffee cup sits next to the laptop, steam rising. The desk chair is empty. The rest of the office is quiet, a few plants, a whiteboard with gestural scribbles in the background. Gently exaggerated proportions, vibrant but restrained color palette with cool-blue office undertones and warm morning light cutting through the window. Wide establishing shot, eye-level. Animated, slightly heightened, never photoreal. Square 1:1. No text, no logos, no readable signage.

2.5x more code is not 2.5x more value

VV
@vcvanessa · now
Jack Dorsey says code output per engineer is up 2.5x. The denominator isn't in the letter. #venturecapital #tech #startups #ai

Jack Dorsey's Q1 shareholder letter has a line that's going to get copy-pasted into a hundred pitch decks: 'Production code changes per engineer are up over 2.5x compared to January.' Block's Q1 results are actually interesting on their own terms — gross profit up 27%, adjusted operating income up 56% to a record $728 million. But the AI productivity stat is the one that's going to travel.

Here's the problem. Every public company CEO is now required to have an AI productivity quote in their shareholder letter. What almost none of them publish is the denominator. A few things that number doesn't tell you:

  • Code changes per engineer measures output velocity, not output quality. Shipping 2.5x more diffs is not the same as shipping 2.5x more correct code. Ask any engineering manager what happens to review load when junior devs have a copilot.
  • January is a convenient baseline. January is when most AI tooling rolled out companywide. Comparing February-April against the month you handed everyone a new instrument is not a productivity study. It's a before-after ad.
  • The cost side is missing. What did the tooling cost? What did headcount cost before versus after? Restructuring charges of $852 million hit Block's GAAP line in Q1. Some of that is legal contingencies, but organizational changes and AI-enabled headcount decisions tend to travel together.

2.5x more code shipped is a real operational signal. It is not a value creation claim. The market will eventually want the other half of the equation.

image prompt only · not rendered