🟢 READ | ⏱ 12 min | 📡 9/10 | 🎯 Software engineers; AI tool users; engineering leads

TL;DR

Based on a survey of ~1,000 engineers (Jan–Feb 2026), Claude Code has surged to #1 in AI tooling adoption just 8 months after launch, displacing GitHub Copilot and Cursor. AI is now fully mainstream: 95% use AI tools weekly, 75% use it for half+ their work, and 56% do 70%+ of their coding with AI. Agent adoption is rising fast (55% regular use), correlating strongly with positive sentiment.

Signal

• Claude Code dominance: Released May 2025, now the most-used tool by wide margin; 75% adoption among tiny companies vs. 35% for GitHub Copilot (large enterprises prefer Copilot due to Microsoft bundling, not user preference).
• Mainstream adoption confirmed: 95% weekly usage; 75% using AI for ≥50% of work; 56% at 70%+ (consistent across company sizes except largest enterprises).
• Model winner: Anthropic Opus/Sonnet 4.5 models mentioned more often than all other models combined for coding; next-gen versions (4.6) will likely extend lead further.
• Agent growth & sentiment: 55% regularly use AI agents; agent users 2x more enthusiastic about AI (61% vs. 36% for non-users); strong correlation between agent use frequency and optimism.
• Tool fragmentation: 70% use 2–4 tools simultaneously; Cursor growing 35% YoY (may overtake Copilot in 6–9 months); OpenAI's Codex already at 60% of Cursor's usage despite being new.

What They're NOT Telling You

The survey captures sentiment among Pragmatic Engineer subscribers—a self-selected audience of engaged, senior engineers already tracking the AI space. Early adopters and Anthropic enthusiasts are likely overrepresented; skeptics or AI-averse teams may be underrepresented. Enterprise adoption data relies on self-reported tool usage rather than hard procurement metrics, so the GitHub Copilot-at-large-companies finding may reflect bundling/lock-in rather than genuine user preference.

Trust Check

Factuality ✅ | Author Authority ✅ | Actionability ✅

Factuality: Data sourced from ~1,000 respondents over 3 weeks (Jan–Feb 2026). Charts and breakdowns are granular and verifiable. No wild claims—findings align with visible market momentum (Claude Code launch timing, Cursor growth, Codex emergence).

Authority: Gergely Orban (Pragmatic Engineer) is a respected voice in engineering culture; survey methodology is transparent (open to subscribers, not cherry-picked). Large sample size (900+) lends credibility.

Actionability: High. Provides clear data on tool adoption rates, model preferences, and feature priorities (agents). Useful for teams deciding tooling strategy, orgs evaluating vendor consolidation, and engineers choosing what to learn next.