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 keynote to engineering leaders says AI coding tools are changing software development practices, with some engineers coordinating five to 10 agent sessions at once. The account also flags concerns about code review, quality and reliability, while stressing that teams and planning remain important. The observations are a snapshot of industry practices, not a comprehensive measure of adoption or productivity.
A report published by The Pragmatic Engineer describes AI coding agents reshaping software work in 2026, with some engineers coordinating five to 10 agent sessions at a time rather than writing every line themselves. The account, drawn from a keynote to engineering leaders and conversations with developers, also warns that code review, quality and reliability are under strain as practices change; it does not provide a representative industry-wide adoption survey.
The report’s author said the keynote at the LDX3 engineering leadership conference in New York was based on visits to AI labs including OpenAI and Anthropic, conversations with companies and engineers, and unpublished data from GitHub, Factory AI and Linear. The conference drew more than 2,000 engineering leaders, according to the account. The source does not publish the underlying datasets or specify their methods in the supplied material.
Several developers described dividing work among concurrent agent sessions. Boris Cherny, identified in the report as the creator of Claude Code, said he uses five terminal sessions locally and runs another five to 10 Claude sessions on the web. Peter Mattis, co-founder of Cockroach Labs, described a typical cognitive limit of five to 10 concurrent agent sessions, sometimes with subagents. Dima Zaytsev, a software engineer at Linear, said he rotates among multiple local worktrees while agents work.
The report identifies possible downsides alongside the new workflow: its author says assumptions about code output have changed, code reviews risk becoming “theatrical,” and quality and reliability are down. These are the report author’s observations, not quantified findings in the material provided. The report also says some fundamentals remain: teams and planning still matter, and non-engineers have not broadly taken over shipping software.
How Agent Work Changes Engineering
When developers supervise several coding agents, their work can shift from writing code directly toward setting tasks, reviewing output and coordinating parallel work. That can alter how teams divide responsibilities, train engineers and measure progress. The accounts in the report suggest this shift is already visible among some experienced developers, but they do not establish how common it is across the industry.
The risks matter because software still needs to be dependable after code is generated. If review becomes less rigorous or teams accept output they do not understand, defects may be harder to catch. The report raises those concerns but provides no quantified error rates or reliability comparison. Its account points to an unsettled practical question: whether new workflows can retain effective oversight as the volume of generated code grows.
As an affiliate, we earn on qualifying purchases.
From Coding Tools to Agent Workflows
Technology workplaces have adapted to earlier shifts, including the spread of the internet, smartphones and cloud computing. The report argues that AI is moving faster and changing more of software development at once. Martin Fowler, an industry veteran quoted in the source, characterized AI’s impact as larger in scale than earlier changes he experienced, including object-oriented programming and agile development.
The report links the current shift to improvements in AI models’ coding abilities around the end of 2025. It also points to a growing use of tools that let engineers assign work to agents and move between tasks while those agents run. These examples capture practices reported by particular developers and teams; they are not evidence that all engineers have stopped coding by hand or that traditional development environments have disappeared.
“Nothing has hit with the magnitude of AI. This is a whole size difference from anything that we’ve faced before.”
— Martin Fowler, industry veteran, as quoted in The Pragmatic Engineer
software development code review software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Adoption and Quality Remain Unmeasured
The source material does not establish what share of engineers now use coding agents, how frequently they use them, or whether the reported multi-agent workflows improve output or productivity. It refers to unpublished data from several companies but supplies no figures, sampling details or comparison baselines. The claimed decline in quality and reliability is also not quantified in the provided account.
It remains unclear how widely teams will adopt agent-centered workflows, how review standards will change, and which kinds of software tasks will still require substantial hands-on coding. The report presents a current snapshot and the author’s interpretation, not a settled forecast for the entire industry.
As an affiliate, we earn on qualifying purchases.
Teams Test New Agent Infrastructure
The report expects cloud-based coding agents and supporting infrastructure, sometimes described as agent “harnesses,” to develop further. It also anticipates that some engineers may read less code directly as they supervise agent work. These are the report author’s expectations, not confirmed outcomes or a published schedule.
The next practical test for companies will be whether they can integrate agent workflows without weakening code review, reliability or team coordination. The source does not identify a specific next milestone or provide a timeline. Its account leaves the industry’s longer-term effects on engineering roles and working practices open.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is changing in software development in 2026?
The report describes some engineers assigning work to several AI coding agents at once and shifting effort toward coordinating and reviewing their output. It does not show that this is standard practice everywhere.
How many AI agents do developers use at once?
Developers quoted in the report described working with roughly five to 10 concurrent sessions. These are personal examples, not an industry average.
Does the report prove AI has reduced software quality?
No. The author raises concerns about quality and reliability, but the supplied material includes no measured error rates or comparative data to prove a decline.
Are engineers no longer writing code by hand?
The report says there are signs that many engineers write less code directly, but it does not provide a representative survey showing that engineers have stopped coding by hand across the industry.
What does the report say remains important?
It says teams and planning still matter, even as AI tools change how coding tasks are carried out. Effective review and reliability remain open challenges.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
