Devs are coding faster. Coding reviews are eating the gains

https://imgproxy.divecdn.com/yz4lQLq7htg2pTtxt2tFwDp7jdAHYyKnFMCul__Waxw/g:ce/rs:fit:770:435/Z3M6Ly9kaXZlc2l0ZS1zdG9yYWdlL2RpdmVpbWFnZS9HZXR0eUltYWdlcy0yMTc4ODQ4OTk3XzIuanBn.webp

An article from

Dive Brief

“The bottleneck has moved from writing code to trusting it,” Purna Doddapaneni, partner at Bain & Co., told CIO Dive.

Published Oct. 2, 2026

This audio is auto-generated. Please let us know if you have feedback.

Dive Brief:

  • AI coding tools are helping developers complete roughly 21% more tasks, but time spent reviewing outputs rose 91%, according to Bain & Co.’s 2026 Global Technology Report released Tuesday. The analyst firm surveyed nearly 300 senior technology leaders for the report.
  • Developers have become orchestrators as they grapple with 47% more workstreams happening simultaneously, a sign that AI is generating work faster than processes can absorb it. Leading AI labs and platforms are investing billions in forward-deployed engineering models to help companies absorb AI into existing workflows.
  • “The bottleneck has moved from writing code to trusting it,” Purna Doddapaneni, partner at Bain and one of the...

Copyright of this story solely belongs to www.ciodive.com. To see the full text click HERE

Read more

https://media.wired.com/photos/6abf96a21c08425a2aa5099d/191:100/w_1280,c_limit/GettyImages-2286465910.jpg

Amazon says it has stopped using NDAs with county officials for data center projects and acknowledges community backlash is leading to data center moratoriums

Sponsor Posts Subquadratic: the LLM built for 12M-token reasoning — SubQ can reason across entire codebases and document sets in one pass with no RAG workarounds. Read how SubQ 1.1 Small holds near-perfect retrieval out to 12M tokens. Introducing Campus: The digital home for educational institutions — Every educational institution needs