State of IT in 2026: Agents Wrote More Code Than Humans
The industry stands at the edge of its biggest transformation in decades. Legendary software architect Martin Fowler compared AI's impact on development to the internet and mobile eras combined: "Nothing has hit this hard before." The past year brought radical shifts: developers have nearly stopped writing code by hand, running packs of 5–10 parallel agents instead. GitHub and Linear published record numbers: in August 2026, agents created more pull requests than humans; since July, agents have been filing more tickets on Linear than people. • IDEs are disappearing as a category — the last major VS Code-based IDE shipped in November 2025, Antigravity 2.0 dropped the editor interface entirely, and Cursor and JetBrains Air are both pivoting toward agentic environments • Every mid-size-and-up company is building its own harness: Stripe (Minions), Uber (Minion), Block (Goose), OpenAI, Google, Meta, Amazon, Dropbox, Spotify • Development now kicks off from Slack with direct commands to agents, and complex migrations are shrinking from years to months
AI-processed from Habr AI; edited by Hamidun News
The strongest shift IT has ever seen
Industry veteran Martin Fowler, a living witness to the OOP, internet, and agile revolutions, argues that what's happening with AI right now has no precedent. Speaking at The Pragmatic Summit, he said: "We always had to explain why object-oriented programming, the internet, and agile mattered. With AI, nobody argues whether it matters. You can't put on blinders and pretend this thing means nothing." The quote captures the scale precisely: these changes aren't unfolding over a year or two — they're happening in real time, before our eyes.
The data backs up the claim. Fowler gathered it for a talk in New York in front of 2,000 engineering leaders, pulling fresh material directly from GitHub, Factory AI, and Linear — numbers that hadn't been published anywhere before.
Ten trends that remade software development
1. Humans have nearly stopped writing code by hand. Even the creator of Ruby on Rails has sparked debate over the death of manual programming. Most engineers have shifted to using AI agents as their primary tool.
2. Parallel agents became the norm. Boris Cherny, creator of Claude Code, runs 5–10 instances of Claude simultaneously in the web version, cycling between them. Peter Mattis, co-founder of Cockroach Labs, described the same pattern: "My cognitive limit is somewhere around 5–10 concurrent agent sessions, but each of those is running a bunch of subagents internally." Dima Zaitsev (an engineer at Linear, previously a colleague at Uber) confirms it: 5–10 local worktrees, switching between them, with one agent working while attention shifts to another.
3. Agents opened more PRs than humans. GitHub reported a ninefold increase in agent-created pull requests over eight months. In August 2026, agents surpassed humans in the number of PRs opened on the platform for the first time. The forecast: agent-created PRs will now always outnumber human ones.
4. Agents create more tickets than humans. Linear added built-in support for agent-created tasks. Since July 2026, agents have been opening more tickets than humans.
5. Skills became the primary tool. Factory AI reported that 83% of its users now use skills — 2.5 times more than in February. Skills aren't just for engineers; everyone uses them, and companies are building repositories of corporate skills.
6. IDEs are disappearing. Steve Yegge of Factory AI declared the end of the IDE era. Cursor was the last major IDE built on a VS Code fork (released November 2025); nothing new has launched since. Antigravity 2.0 dropped the IDE interface entirely, Codex moved to a desktop app with no IDE, and JetBrains launched JetBrains Air focused on agentic environments. Kent Beck explained: "The context has changed. The IDE is turning into an interface for dialogue and for watching an agent work, not a code editor."
7. Every company is building its own harness. From mid-size companies on up, nearly everyone has one: Stripe (Minions), Uber (Minion), Block (Goose), Shopify (River), Google (Agent Smith), Meta (Devmate), Amazon (Kiro + Crew), Dropbox (Nova), Spotify (Honk), DoorDash (Flux), Grab (LLM-Kit), WorkOS (Horizon), Hubspot (Crucible), Monzo (Agent Chip), Sierra (Pinecone), Harvey (Spectre), Browserbase (bb).
8. Development now starts in Slack. OpenAI types "@Codex, implement this," Anthropic types "@Claude, do this." Startups are hooking their own coding agents into Slack. The approach that works best is deep integration: a Slack agent triggers a cloud agent to write the actual code.
9. Agentic software factories are being built everywhere. OpenAI showed how this works in practice: AI gets embedded into existing systems (CI/CD, deployment), and new tools emerge — agentic code review, agentic deployment systems, Perf Factory.
10. Migrations now take weeks, not years. Anthropic: porting Bun from Zig to Rust took one and a half engineer-years of work. OpenAI: migrating its API layer from Python to Rust took about five months instead of several years. Airbnb: migrated 3,500 end-to-end test files from Enzyme to React Testing Library. Asana: migrated 4,000 files from Enzyme to React Testing Library in a few weeks, despite the work being originally estimated at five years.
The industry is moving into a new order of speed and scale. Martin Fowler is right: this is a fundamentally different kind of change.
- Oct 27, 2026OpenAI Says Its AI Disproved a 30-Year-Old Theoretical Computer Science Conjecture
- Oct 27, 2026Anthropic Files S-1 with the SEC: IPO Under Ticker ANT and $16.3 Billion in Six Months
- Oct 27, 2026OpenAI Releases GPT-6.1 Terra: Astra-Level Computer Control at 1/10 the Price
- Oct 27, 2026OpenAI Releases GPT-6.1 Terra: Tokens Cost Half as Much, Hallucinations Drop From 88% to 38%
Key Questions Answered
What are the key details on The strongest shift IT has ever seen?
Industry veteran Martin Fowler, a living witness to the OOP, internet, and agile revolutions, argues that what's happening with AI right now has no precedent. Speaking at The Pragmatic Summit, he said: "We always had to explain why object-oriented programming, the internet, and agile mattered. With AI, nobody argues whether it matters. You can't put on blinders and pretend this thing means nothing." The quote captures the scale precisely: these changes aren't unfolding over a year or two — they're happening in real time, before our eyes. The data backs up the claim. Fowler
What are the key details on Ten trends that remade software development?
1. Humans have nearly stopped writing code by hand. Even the creator of Ruby on Rails has sparked debate over the death of manual programming. Most engineers have shifted to using AI agents as their primary tool. 2. Parallel agents became the norm. Boris Cherny, creator of Claude Code, runs 5–10 instances of Claude simultaneously in the web version, cycling between them. Peter Mattis, co-founder of Cockroach Labs, described the same pattern: "My cognitive limit is somewhere around 5–10 concurrent agent sessions, but each of those is running a bunch of subagents internally."
Delegate Deep Research, Coding & Workflows to an AI Agent
Deploy autonomous multi-step execution with NomadNet Agent (agent.nomadnet.ai) — automated web research, data synthesis, code generation, and complex task orchestration.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.