Urgent.News

What's breaking now, across hundreds of outlets.

AI

AI Coding Agents Raised Commits 240%. Releases Rose Only 30%

Autonomous coding agents lifted commit volume 240% across more than 500,000 GitHub developers, yet software releases rose only 30% (Source: National Bureau of Economic Research, 2026). The Philippines is buying this capability fast: 20.7% of working-age Filipinos used a generative AI product in the second quarter of 2026, above the 18.8% global average (Source: Microsoft AI Economy Institute,…

Autonomous coding agents led to a 240% increase in code commits across over half a million GitHub users, yet software releases only rose by 30% in the same period (Source: National Bureau of Economic Research, 2026). The Philippines is quickly adopting this technology, with 20.7% of working-age individuals using a generative AI product in Q2 2026, exceeding the global average of 18.8% (Source: Microsoft AI Economy Institute, 2026).

The Philippines' Department of Science and Technology invested P2.6 billion in its National Artificial Intelligence Center for Research and Innovation in February 2026 (Source: Rappler, 2026).

Generation of AI coding assistance is no longer a limitation, and the focus has shifted to problem-solving downstream. The study discovered that the 240% increase in commits diminishes as you move down the production hierarchy, falling to 80% for projects and 30% for releases (Source: National Bureau of Economic Research, 2026).

This phenomenon is attributed to the "weak-link hypothesis," where the elasticity of substitution between AI and human effort is estimated at 0.23 (Source: National Bureau of Economic Research, 2026). Despite the increase in new app development across major software marketplaces, there was no corresponding rise in total usage (Source: National Bureau of Economic Research, 2026).

Controlled experiments at Microsoft, Accenture, and a Fortune 100 company involving 4,867 developers showed a 26.08% increase in completed tasks with an AI coding assistant (standard error of 10.3%); the effect was more pronounced among less experienced developers (Source: Management Science, 2026). Microsoft's employee trial found that AI was expected to make developers 24% faster, but they were 19% slower than anticipated (Source: METR, 2025).

Similarly, a follow-up study by METR in February 2026 indicated speedups of -18% for experienced developers and -4% for newcomers, both with confidence intervals crossing zero, indicating weak evidence (Source: METR, 2026).

Using over 30 million GitHub commits, a classifier estimated that AI writes 29% of Python functions in the US, with quarterly code contributions increasing by 3.6%. Experienced programmers accounted for the majority of this growth (Source: Science, 2026).

A March 2026 DORA analysis of 1,110 Google software engineer responses revealed that time saved through code generation often gets redirected to auditing and verification, linking higher AI adoption to both increased throughput and higher instability in the software (Source: DORA, 2026). Similarly, a 2025 DORA report found that 90% of technology professionals use AI at work, with 80% believing it increased their productivity, but 30% did not trust AI-generated code (Source: DORA, 2025).

Rewritten by urgent.news; may contain errors — check the original.

Read the original at dev.to →

Report a problem with this story

More in AI

From task.md to Orca: How I Streamlined My Claude Code and Codex Workflow

In a previous article, How I Use Claude Code and Codex Together to Build .NET Features Faster , I described a workflow in which Claude Code and codex linked together to achieve my codings tasks.

  • Orca replaces task.md in Claude Code workflow
  • Orca uses repository as project exploration space
  • Orca streamlines workflow by reducing setup time

Reconciling two Excel workbooks offline with a local Windows AI assistant

What should a local language model do when a user asks, “What changed in these two spreadsheets?” A plausible answer is easy to produce.

  • Local language model compares two Excel workbooks offline
  • User selects original file, revised file, and key column on each sheet
  • Model retrieves price data directly from spreadsheet cells

More from Saturday 21 November →