TL;DR
Get the latest gadgets delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
A report based on a 2026 engineering leadership keynote says AI tools are changing how software teams write and review code, with many engineers coordinating several coding agents at once. The account also flags weaker quality and reliability and says it is not yet clear how broadly these practices apply across the industry.
AI coding agents are changing how software engineers work, according to a 2026 industry report from The Pragmatic Engineer, which draws on a keynote delivered at the LDX3 engineering leadership conference in New York. The report describes engineers increasingly managing five to 10 agent sessions in parallel, while warning that code quality, reliability and review practices have not kept pace with the shift.
The report’s author says the keynote was informed by visits to AI labs including OpenAI and Anthropic, conversations with companies including Ramp and Uber, and unpublished data from GitHub, Factory AI and Linear. The conference drew more than 2,000 engineering leaders and senior technical staff, according to the report. Its observations cover AI labs, venture-backed startups and large technology companies, but do not establish that every organization has adopted the same practices.
Several engineers quoted in the report describe delegating coding tasks to multiple AI agents, then moving among sessions to check results or continue other work. Claude Code creator Boris Cherny said he uses five local terminal sessions and runs another five to 10 Claude sessions on the web. Linear software engineer Dima Zaytsev described rotating among five to 10 local worktrees while agents work. These accounts illustrate individual workflows, rather than a measured industry-wide average.
The report says assumptions about code output have changed, and that code reviews can become “theatrical” when teams do not adequately verify generated work. It also identifies lower quality and reliability as problems emerging alongside faster production. The author says teams and planning still matter, and reports no general shift to non-engineers shipping code. These are the report’s assessments; it does not provide a quantified industry-wide measure of productivity or defect rates.
Productivity Gains Bring Review Risks
If engineers can delegate more implementation work to agents, software teams may be able to attempt more tasks at once and change how they allocate time. The accounts in the report suggest a move from writing each line directly toward orchestrating, testing and reviewing agent output. That shift could affect hiring, engineering tools and the skills companies value, though the report does not quantify those effects.
The warning about reliability matters because more code produced does not automatically mean more dependable software. When developers supervise several agents, teams need workable methods to check that code meets requirements and behaves safely. The report’s description of review becoming performative points to a risk that faster output can outpace careful verification. Whether organizations can raise output without raising defects remains unsettled.
As an affiliate, we earn on qualifying purchases.
From Coding Tools to Agent Workflows
The report places the current change alongside earlier shifts such as the internet, smartphones, cloud computing and new programming languages. It argues that AI’s influence is unusually large and fast-moving, while acknowledging that parts of software development remain familiar: teams and planning still count. The central change is not simply adopting another tool, but delegating increasing portions of implementation to systems that can work across tasks.
Martin Fowler, a software development expert, described AI’s impact as a “whole size difference” from earlier changes in a statement quoted by The Pragmatic Engineer. The report also links the acceleration to improved coding models near the end of 2025. It presents this as a turning point in the author’s assessment, rather than as a finding established through a controlled study of the whole industry.
“Nothing has hit with the magnitude of AI. This is a whole size difference from anything that we’ve faced before.”
— Martin Fowler, software development expert, quoted by The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
How Widespread Are These Practices?
The report offers examples and observations, but its source material does not provide a representative survey showing how many engineers use coding agents, how often they do so, or how adoption varies by company size and sector. The cited workflow accounts come from individual engineers and should not be treated as an industry-wide norm.
It is also unclear how large the reported quality and reliability problems are, how they compare with pre-AI development, or whether particular review methods can reduce them. The report references unpublished data from several companies but does not state the figures in the supplied material. Claims about productivity and the pace of change should therefore be read as reported observations, not settled measurements.
developer workstation with multiple terminal sessions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Tracking Tools, Quality and Adoption
The report expects cloud-based coding agents and the systems that coordinate them, often called harnesses, to gain attention. It also anticipates companies developing new AI infrastructure and engineers spending less time reading code directly. These are the report author’s expectations, not confirmed outcomes or announced industry-wide plans.
The next useful evidence will be clearer measures of adoption, output and software defects, alongside accounts of how teams review agent-generated code. Until such data is available, the key question is whether organizations can make parallel agent workflows dependable at scale—not only whether they can produce code faster.
AI-powered code quality analysis tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the main change described in the report?
The report says some engineers are shifting from writing code line by line to directing several AI coding agents and checking their output.
Does the report show that most engineers use multiple agents?
No. It gives examples from individual engineers but does not provide a representative adoption survey. The prevalence of these workflows across the industry is not established.
What concerns does the report raise?
It identifies possible declines in code quality and reliability and says reviews can become performative if generated work is not properly checked. The supplied material does not quantify these problems.
What does the report say will happen next?
The author expects growth in cloud coding agents, agent-coordination tools and AI infrastructure. These are predictions; their scale and timing remain uncertain.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
