Slide contents

1. 1. From AI-Native Development to AI-Native Platform

   How agentic workload changes development platforms — using GitHub as an example

2. 2. What This Article Is About

   Load unit + platform

   Parts 1-2: GitHub platform case.

   Part 3: broken load economics.

   Part 4: AI-native platform failure.

   Parts 5-6: lessons for orgs.

3. 3. 01. GitHub as AI-Native Platform

   GitHub is a control plane for people and agents

4. 4. GitHub 2025–2026

   AI is first-class load

   +36M in 2025 — 180M+ developers — 80% tried Copilot in week one.

   ×2 in <2 years — 4.3M AI repositories — AI repos: 230 per minute.

   5.5M issues/month — 986M commits in 2025 — Copilot correlates with 8-15% growth.

5. 5. Task Reaches an Agent

6. 6. GitHub Strengthens Agent Mode

7. 7. 02. What the Numbers Show

   Actions, code review, usage metrics — three signals of new pressure

8. 8. AI-load signals

   AI is first-class load

   11.5B minutes · +35% YoY — GitHub Actions — 71M jobs/day; agents run on Actions.

   60M reviews · ×10 growth — Copilot Code Review — >1/5 reviews; 71% actionable feedback.

   First-class traffic — AI-native telemetry — Copilot PRs, cloud agents, CLI tokens.

9. 9. Review Quality vs Computational Depth

10. 10. 03. Why Old Architectural Assumptions Break

    Agentic fan-out: one issue spawns a tree of platform operations

11. 11. Old vs New Unit of Load

12. 12. 04. AI-Native Platform Failure

    New load meets old architecture

13. 13. Incident: Feb 9, 2026

14. 14. April Incidents

15. 15. Shared Failures Spread Wider

16. 16. 05. For Large Engineering Orgs

    Process-wise — integrate; infrastructure-wise — isolate

17. 17. Process Integration, Infrastructure Isolation

18. 18. DORA: AI Amplifies Platform Properties

    Discipline becomes visible

    DORA 2024: better docs, quality, review speed.

    Cost: lower throughput and stability.

    DORA 2025: throughput vs stability.

    GitHub shows this dynamic.

19. 19. 06. GitHub Outlook and Conclusions

    The right problem class — the right category of actions

20. 20. GitHub Response: Platform Work

21. 21. Moderately Positive Medium-Term Outlook

    Right direction; pressure grows

    Fix substrate while load grows.

    Multi-provider and mobile agents.

    Runtime SDK and remote CLI sessions.

    Short-term cautious; medium-term positive.

22. 22. Key Takeaways

23. 23. References and Materials

    Core sources

    Classic PDLC to AI-native article.

    GitHub Blog, GitHub Status and Copilot updates.

    GitHub Actions, Code Review, org metrics.

    DORA 2024 report.

24. 24. Thank You!

    AI-Native Platform

    For more materials on this topic, visit the "Book Cube" channel — all links are collected there

    Alexander Polomodov, Technical Director & Fellow, T-Technologies

    @Book_Cube