Copilot Goes Mainstream
Core finding: 80% of new developers use Copilot within their first week—AI-assisted coding is now the default entry point.
The most striking signal from Octoverse 2025 isn't a gradual adoption curve—it's a step function. 80% of new developers on GitHub now use Copilot within their first week on the platform. For the next generation of developers, AI assistance isn't an add-on; it's table stakes.
This shift accelerated dramatically after GitHub launched Copilot Free in December 2024, removing the paywall barrier for individual developers. The timing coincided with record developer sign-ups—over 36 million new developers joined GitHub in 2025, roughly one per second.
| Copilot Milestone | Impact |
|---|---|
| Copilot Free launch | Dec 2024 |
| New developer Copilot usage | 80% in first week |
| Code review effectiveness | 72.6% report improvement |
| Coding agent PRs authored | 1M+ (May-Sept 2025) |
The implication is profound: developers entering the workforce today have never known professional coding without AI assistance. Their expectations, workflows, and skill development are being shaped by a fundamentally different paradigm than developers who came before them.
New Developer Copilot Adoption
80% of new developers use Copilot within their first week on GitHub
The LLM SDK Explosion
Core finding: LLM SDK adoption grew 178% YoY—developers are building AI into everything.
Beyond using AI tools, developers are increasingly building with AI. The number of repositories integrating LLM SDKs—libraries for connecting to models like GPT, Claude, Llama, and others—has exploded:
- 1.1 million+ repositories now use an LLM SDK
- 693,867 new LLM SDK repos created in the past 12 months
- 178% year-over-year growth (August 2024 to August 2025)
This isn't just AI researchers and ML engineers. The SDK adoption pattern shows AI integration spreading across application categories: web apps, mobile apps, internal tools, automation scripts, and more. LLM capabilities are becoming a standard component in the developer toolkit, like databases or authentication.
The growth rate—nearly tripling in a single year—suggests we're still in the early steep part of the adoption curve, not approaching saturation.
LLM SDK Repository Explosion
Repositories integrating LLM SDKs grew 178% year-over-year
AI Repository Landscape
Core finding: 4.3 million AI-related repositories now exist on GitHub—nearly doubled since 2023.
The sheer scale of AI development activity on GitHub has reached a new order of magnitude:
- 4.3 million AI-related repositories total
- Nearly doubled since 2023
- AI projects among the fastest-growing categories on the platform
What's inside these repositories reveals the breadth of AI development:
| Category | Activity Pattern |
|---|---|
| Model fine-tuning | Custom LLM adaptations |
| Inference optimization | vLLM, llama.cpp popularity |
| AI applications | Chatbots, agents, assistants |
| ML pipelines | Training and evaluation |
| AI tooling | Frameworks and utilities |
The repository growth reflects both the democratization of AI development (more developers can build AI applications) and the genuine usefulness of AI capabilities (more applications benefit from AI features).
TypeScript's AI-Driven Rise
Core finding: TypeScript overtook Python and JavaScript as #1 language—the biggest language shift in over a decade, driven by AI's need for type safety.
In August 2025, TypeScript achieved something unprecedented: it became the most-used language on GitHub by contributor count, surpassing both Python and JavaScript. This represents a 66% year-over-year surge—the largest language movement in more than a decade.
The connection to AI is direct. TypeScript's type system provides several advantages for AI-assisted development:
Why AI favors typed languages:
- Better code suggestions — Type information helps AI generate more accurate completions
- Safer refactoring — AI-suggested changes can be validated against type contracts
- Clearer intent — Types communicate developer intent that AI can leverage
- Fewer runtime errors — Type checking catches mistakes before AI-generated code runs
| Language | YoY Contributor Growth | Rank Change |
|---|---|---|
| TypeScript | +66% (+1M contributors) | → #1 |
| Python | +48.7% (+850K contributors) | Held #2 |
| JavaScript | Moderate growth | → #3 |
Python remains dominant for AI/ML workloads specifically, but the broader ecosystem shift toward TypeScript reflects how AI assistance is reshaping language preferences across all development contexts.
TypeScript's AI-Driven Rise to #1
Year-over-year contributor growth — biggest language shift in over a decade
Why AI favors typed languages:
Agentic Coding Arrives
Core finding: The Copilot coding agent authored over 1 million pull requests in just five months.
2025 marks the emergence of AI agents as active participants in the development process—not just suggesting code, but authoring complete contributions:
- 1 million+ pull requests authored by the Copilot coding agent
- Timeframe: May to September 2025 (5 months)
- Following the coding agent preview launch in March 2025
This represents a qualitative shift from AI-as-assistant to AI-as-contributor. The agent doesn't just complete lines of code; it can understand issues, plan implementations, write code across multiple files, and submit coherent pull requests for human review.
Early patterns show agents particularly effective for:
- Bug fixes with clear reproduction steps
- Dependency updates and maintenance
- Test coverage expansion
- Documentation improvements
- Refactoring with well-defined scope
The 1 million PR milestone in five months suggests we're seeing just the beginning. GitHub's report notes: "We expect far greater activity in months and years ahead."
Copilot Coding Agent Activity
Over 1 million pull requests authored in just 5 months (May-Sept 2025)
Where agents excel:
Open Source AI Dominance
Core finding: AI projects grow contributors 150% YoY on average—3x the typical open source project.
AI hasn't just created new repositories; it's created a new gravity well for developer attention and contribution:
- 6 of the top 10 fastest-growing projects on GitHub are AI-focused
- 6 of the top 10 projects by contributor count are AI-centric
- AI projects average +150% contributor growth YoY
- Typical open source project median: ~50% contributor growth
Leading AI open source projects:
| Project | Focus | Growth Pattern |
|---|---|---|
| vllm-project/vllm | High-throughput inference | Top 10 fastest |
| huggingface/transformers | Model library | Top by contributors |
| ollama/ollama | Local model running | Top 10 fastest |
| llama.cpp | Lightweight inference | Top 10 fastest |
The contributor intensity around AI projects reflects both the technical complexity (many hands needed) and the strategic importance (companies and individuals investing heavily in AI capabilities).
AI Repository Landscape
4.3 million AI-related repositories — nearly doubled since 2023
Leading AI Open Source Projects
The New Developer Journey
Core finding: 20% of the most popular projects for first-time contributors are now AI-related.
The developer onboarding experience has fundamentally shifted. New developers aren't just learning to code—they're learning to code with AI, and often learning by contributing to AI projects:
- 20% of most popular newcomer projects are AI-focused
- March 2025: 255,000 first-time open source contributors (record month)
- AI projects provide accessible contribution opportunities
This creates a feedback loop: new developers use Copilot → gain interest in AI → contribute to AI projects → deepen AI skills → build AI applications.
The new developer stack:
- IDE with Copilot integration (VS Code dominant)
- LLM SDK in early projects
- AI-assisted debugging and code review
- Contribution to AI open source
- Building AI-powered applications
Developers entering the field today will have AI fluency as a baseline assumption, not a specialized skill.
AI Infrastructure Winners
Core finding: Inference optimization and local model running dominate the fastest-growing AI projects.
The specific projects gaining the most traction reveal where AI development energy is concentrated:
Inference layer:
- vLLM — High-throughput, memory-efficient serving
- llama.cpp — Lightweight CPU/GPU inference
- ollama — Local model management and running
Model ecosystem:
- Hugging Face transformers — Model hub and library
- LangChain — LLM application framework
- Various fine-tuning tools — Custom model adaptation
The pattern suggests developers are moving past "can we use AI?" to "how do we deploy AI efficiently?" Infrastructure for serving, optimizing, and managing models is now as important as the models themselves.
Development environment signals:
- Jupyter notebooks: 2.4M repositories (+75% YoY)
- Dockerfiles: 1.9M repositories (+120% YoY)
The notebook growth reflects AI/ML experimentation, while Docker growth indicates production deployment patterns—developers are moving AI from notebooks to containers.
Productivity Signals
Core finding: 72.6% of Copilot code review users report improved effectiveness.
GitHub's data provides several productivity indicators around AI-assisted development:
Copilot code review:
- 72.6% of users report improved effectiveness
- Autofix resolving OWASP Top 10 issues across thousands of repos monthly
- Critical vulnerability fix time: 30% faster with automation
Activity trends (2025 vs 2024):
- Pull requests merged: +23% YoY (43.2M monthly average)
- Commits: +25% YoY (986M total, 100M in August alone)
- Code pushes: +26% YoY (82M monthly average)
Interesting signals:
- Comments on commits: -27% (less back-and-forth needed?)
- Comments on issues/PRs: Essentially flat (+0.35%)
The decline in commit comments alongside increased PR velocity could suggest AI-assisted code requires less revision discussion—though alternative explanations exist.
What's Next
Core finding: The data points to agents, typed languages, and AI-native workflows becoming standard.
GitHub's Octoverse data suggests several trajectories for AI in development:
Near-term (2025-2026):
- Agent capabilities expanding beyond current coding agent scope
- Copilot integration deepening across GitHub workflows
- TypeScript and typed language adoption continuing to climb
- LLM SDK integration becoming as common as REST API usage
Medium-term signals:
- AI-first development environments becoming default
- New developer expectations reshaping tools and practices
- Open source AI infrastructure maturing toward production-grade
- Geographic shift—India's developer population (21.9M → projected 57.5M by 2030) increasingly AI-fluent
Open questions:
- How will AI agents change code review practices?
- Will typed languages continue gaining share from dynamic alternatives?
- What new security challenges emerge from AI-generated code at scale?
- How do developer skills evolve when AI handles routine tasks?
The Octoverse data captures a moment of transformation. The 80% first-week Copilot adoption rate among new developers isn't just a metric—it's a generational marker. Developers entering the field now will never know professional coding without AI assistance, and that shapes everything that follows.
This analysis is based on GitHub's Octoverse 2025 report, drawing on activity data from 180 million+ developers and 630 million repositories. The full report is available at octoverse.github.com.
Source
GitHub - GitHub Octoverse 2025: The AI Developer Revolution. October 2025.
Available at: https://octoverse.github.com/