TL;DR
The Pragmatic Engineer’s 2026 industry report describes AI coding tools shifting how many software engineers work, including the use of multiple agents in parallel. The report also identifies concerns about code quality and reviews, while arguing that teams and planning still matter; the scale of adoption and its effects remain unsettled.
AI coding agents are changing how software engineers work, with some experienced developers now running several agents at once instead of writing each line by hand, according to a 2026 industry report from The Pragmatic Engineer. The report, based on conference observations, interviews and data shared by technology companies, also flags concerns about code quality, reliability and the substance of code reviews as the pace of adoption accelerates.
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The report’s author says coding models improved substantially toward the end of 2025, helping make AI-assisted development a wider workplace trend in 2026. In the account, many engineers increasingly delegate coding tasks to agents and oversee multiple work sessions in parallel. That is a description of observed practices, not a measured estimate of how many engineers have stopped coding by hand across the industry.
Examples come from developers interviewed by the author. Claude Code creator Boris Cherny described using five local terminal sessions and another five to 10 Claude sessions on the web. Peter Mattis, co-founder of Cockroach Labs, said his cognitive capacity is often around five to 10 concurrent agent sessions, sometimes with subagents. Dima Zaytsev, a software engineer at Linear, described rotating among five to 10 local worktrees while agents work on separate tasks.
The report also lists potential downsides: assumptions about generated code may no longer hold, code reviews can become performative, and quality and reliability may decline. Those are the author’s assessments; the supplied material does not provide quantified industry-wide measurements for these effects. The report says some practices have not changed: teams and planning remain important, and it does not claim non-engineers are now routinely shipping production code.
AI Changes the Shape of Coding Work
The shift matters because delegating implementation to agents changes what engineers spend time doing. The examples point toward more work coordinating tasks, checking outputs and moving between parallel sessions. If that approach spreads, companies may need to revise how they plan software projects, assess engineering productivity and assign responsibility for defects.
The report’s concerns about review and reliability highlight a trade-off: producing code more quickly does not by itself establish that the software is correct or maintainable. Teams still need ways to test work and judge whether an agent’s output meets requirements. The source material does not establish that AI tools have caused a broad decline in software quality, but it identifies that risk as a live issue for engineering leaders.
Popular right now

Roborock Qrevo S Pro Robot Vacuum and Mop with 18,500 Pa Suction

KitchenAid Artisan 5-Qt Tilt-Head Stand Mixer, Almond Cream

Dyson V8 Cyclone Cordless Vacuum
As an affiliate, we earn on qualifying purchases.
From Coding Models to Agent Workflows
The author places the current change alongside earlier shifts in software, including the spread of the internet, smartphones, cloud computing and new programming languages and frameworks. The report argues that AI’s pace and scale stand out even to people who have worked through those earlier changes. That is an informed perspective from the report’s author and interviewees, rather than a statistical comparison across technology eras.
The snapshot draws on the author’s keynote at the LDX3 engineering leadership conference in New York, which the source says drew more than 2,000 engineering leaders and senior technical staff. The author also cites visits to OpenAI and Anthropic, conversations with startups and technology companies, and data described as unpublished from GitHub, Factory AI and Linear. The supplied material does not give the underlying datasets or their methods, so those references cannot support independent conclusions about industry-wide adoption rates.
“Nothing has hit with the magnitude of AI. This is a whole size difference from anything that we’ve faced before.”
— Martin Fowler, software engineer and author, speaking at The Pragmatic Summit
Adoption and Quality Remain Unmeasured
The report does not establish what share of engineers or companies now rely on coding agents, or how that share differs by role, company size or type of software. The examples are individual accounts, and the material provides no sampling method for treating them as representative. It also does not publish the cited company datasets, making their scope and results unclear.
The extent of any change in code quality, reliability or review standards is also unresolved in the supplied material. The report identifies these as problems or risks, but gives no industry-wide measurements or comparisons that would show their prevalence or connect them directly to AI use.
More Agent Infrastructure Is Emerging
The report expects cloud-based coding agents and the software “harnesses” around them to develop further, along with new infrastructure built specifically for AI-assisted engineering. These are forecasts in the report, not confirmed outcomes. It also suggests that engineers may spend less time reading code directly, a prediction whose extent and timing remain uncertain.
The next useful evidence will be clearer data on how organizations use agents, whether teams change their review and testing practices, and what happens to quality and reliability over time. The report provides a snapshot of current practices and concerns, but does not resolve those questions.
Key Questions
What is changing in software engineering in 2026?
The report describes engineers using AI coding agents to complete tasks and working across several agent sessions at once. It presents this as an emerging practice, not a measured description of every engineering team.
Are engineers no longer writing code by hand?
The report says there are signs that many engineers write less code by hand, but the supplied material gives no industry-wide adoption figure. Individual examples should not be taken as proof that hand-coding has ended.
Does the report show that AI has reduced software quality?
No quantified industry-wide result is provided. The author identifies quality and reliability as concerns, but the material does not establish their prevalence or prove that AI caused a decline.
What does the report say is staying the same?
It says teams and planning remain important. It also does not claim that non-engineers are routinely shipping production code as a result of AI tools.
Source: rss
