How Tech Giants Like Uber, Stripe and Shopify Are Automating Code With AI Factories

·1 views

Major tech companies are deploying multi-agent AI pipelines — dubbed 'software factories' — to automate the bulk of their software development workflows. Stripe merges over 1,000 AI-written pull requests weekly, Shopify's bot co-authors one in eight merged PRs, and Uber attributes more than 70% of its pull requests to AI agents as of mid-2026. Unlike standard AI coding assistants that require a human to prompt each step, these factories operate as end-to-end pipelines: an event like a bug report triggers agents that triage, implement, test, and review code before producing a pull request for human approval. Vercel built an open-source factory to tackle a backlog of over 1,000 issues and 800 open PRs, while Cloudflare and Astro's triagebot-action cut Astro's open issue count from over 200 to around 30. Human oversight remains a consistent feature across these systems, with engineers retaining final merge authority even as autonomous agents handle an increasing share of routine engineering tasks.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Seven Common Reasons RAG Chatbots Fail in Production and How to Fix Them

RAG chatbots that perform well in demos often break down in production due to architectural flaws rather than model limitations, according to a technical guide published on DEV Community. The core problem is that most systems are built as a single monolithic LLM call with a static vector store, which cannot handle unpredictable real-world queries or constantly changing data. A modular architecture is recommended, separating concerns into distinct layers for orchestration, retrieval, generation, and output validation. The guide also highlights that retrieval should not be limited to vector databases alone, as production systems often need to query SQL databases, REST APIs, and file systems simultaneously. The article provides TypeScript-based blueprints to help developers build more reliable and scalable RAG systems suitable for use cases like customer support and internal knowledge assistants.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

HTTP/3 Is Not a Silver Bullet for Unity Game Networking, Developers Warned

A technical guide targeting Unity 6.3–6.5 developers cautions against assuming HTTP/3 automatically improves game performance. The article breaks down three protocol tiers — HTTP/1.1, HTTP/2, and HTTP/3 — across distinct use cases including REST APIs, asset delivery, and real-time networking. HTTP/2 is recommended as the practical baseline, with HTTP/3 considered only when real-device testing on lossy mobile links demonstrates measurable gains. Real-time game state synchronization should rely on dedicated solutions like Photon or Unity Transport rather than HTTP protocol upgrades. Developers are also warned about HTTP/3 edge cases, including UDP port restrictions on some networks and replay risks with QUIC's 0-RTT feature.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Free In-Browser SQL Practice Tool Targeting Industry Interview Gaps

A developer created dataDrill after failing a MAANG technical interview that required advanced SQL skills not taught in his college database course. The platform runs a real PostgreSQL engine entirely in the browser via WebAssembly, requiring no installation or configuration. It features over 90 SQL questions sourced from technical screens at companies like Meta, Google, Stripe, and Netflix, focusing on window functions, self-joins, and date math. Each challenge includes line-by-line solution explanations to help users understand syntax rather than just copying answers. Several difficult questions are available for free, with the tool aimed at CS graduates and junior developers preparing for data engineering or analyst roles.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Compares Building the Same Security App on macOS and Linux

A developer built the same background security application on both macOS and Linux, documenting the contrasting experiences. On macOS, they found a tightly controlled environment where Apple's code signing, notarization, and entitlements system restricts what apps can do but builds user trust automatically. Linux offered far greater freedom — allowing low-level system access via tools like fanotify and nftables — but required manual configuration and placed full responsibility for security on the developer. The author concludes that the core difference lies not in available tools but in how each platform assigns responsibility for trust and safety. They note that macOS holds official UNIX 03 certification, while Linux, despite sharing Unix conventions, has never been formally certified.

0 comments Read more at DEV Community