Wix Engineering

Millions of users. Critical production databases. Zero downtime. 🚩 No room for mistakes. At @Wix, shared DB clusters created a massive reliability challenge: a single problematic application could impact many others. In this blog post, Bar Shauli explains how she, and the Wix Data Infrastructure team, built an internal migration service using Kafka and ProxySQL on top of MySQL - to safely move databases between clusters at huge scale-transparently, fully online, with ✅ zero downtime. Learn how mega-scale database migrations actually work behind the scenes, including live traffic 🚦 cutovers, streaming, partitioning challenges, and more. Read the full story here: wix.engineering/post/how-we-bu…

Wix ran 250 controlled evaluations comparing AI agent performance with documentation versus curated skills, finding well-maintained docs often outperform purpose-built skills. The logic seems bulletproof: if you want an AI agent to use your platform, you shouldn’t just give it raw documentation. You should give it a "skill", a curated, condensed, and optimized guide. This will allow the agent to perform tasks on your platform better than if they have to trawl through your docs. Skills are intuitive and trendy, but do they really provide agents with an edge over just using the docs and, if so, in what cases? At @WixEng , we decided to question the hype and start measuring. We ran 250 controlled evaluations comparing how AI agents perform tasks using standard docs, AI-optimized docs, and purpose-built skills. The results were surprising and they challenged our entire strategy for the AI-native developer experience. #AI #skills #agents #wix #wixengineering wix.engineering/post/we-ran-25…

Most AI coding workflows eventually hit the same 🚩 problem: the Context Wall. The bigger the project gets - the more the AI loses track. In this talk, @yoavabrahami introduces the "Design-Log Methodology" - a simple but powerful shift from endless prompting to version-controlled architectural thinking. Instead of constantly fighting context drift, you’ll learn how to turn AI into a real architectural partner - with: ✅ more predictable results ✅ faster delivery ✅ and dramatically less back-and-forth. Watch:

Full house 🔥 right now at the @Wix Campus! Our “AI-Native Development: From Understanding Code to Fixing It Autonomously” meetup with @guy_bary and @IsraelZablianov - is live with many developers diving into how AI is transforming software engineering: ✅ AI-powered code research engines ✅ Autonomous production investigations ✅ and systems that can actually debug and fix software The full sessions will be shared soon!

Getting ready 🔥

בזמן שחברות בודדות מגייסות ג׳וניורים, וויקס פותחת תכנית הכשרת סטודנטים ומראה שוב כמה היא חשובה לאקוסיסטם שלנו. כל הכבוד לאבירן והצוות @aviranm 👏 סטודנטים למדמ״ח - אל תפספסו.

סטודנטים למחשבים - ההרשמה לתוכנית הסטודנטים של וויקס נפתחה ל 48 שעות בלבד. סטודנטים למדעי המחשב והנדסת תוכנה המסיימים ב 2028 לפחות, זו ההזדמנות שלכם. הגישו מועמדות ושתפו עם חברים, זו תוכנית הסטודנטים היחידה שתיפתח השנה שתכשיר אתכם להיות מהנדסי תוכנה בעולם ה AI. הגישו עכשיו: wixenter.com

Tomorrow 👇

🔥 Meet our next speaker for AI JavaScript fwdays’26: Danylo Kolesnikov, Engineering Team Lead at Wix. 📌 “Remote Agents in Production: From Jira to PR Without a Developer” How Wix builds AI agents that turn Jira & Slack tasks into ready-made PRs. 🎟️ bit.ly/4vuZXiL

What does a web framework look like if it’s designed for the AI Developer? We build web frameworks (React, Vue, Next.js) for humans. Then we spend all our time teaching AI how to "think" like a human dev so it can use them. We create "Skills". Train models. Create "guides". Great "teams" of AI agents to mimic human teams. This feels like a middle-step. If we stop optimizing for "humans", what becomes the new priority? I’d love to hear your wildest ideas on this. How does the stack change when the primary coder isn't human?

What started 11 months ago as a small open-source 🐙 project for searching code across GitHub evolved into a powerful AI research engine used to investigate complex production issues at scale. Today, Octocode - created by @guy_bary - powers autonomous agentic research workflows at @Wix, helping engineers analyze complex bugs and accelerate issue resolution. At our upcoming 👋 meetup (May 11), Guy will showcase how we integrated Octocode into a multi-agent research pipeline - demonstrating how autonomous research flows can investigate tickets, analyze production issues, and dramatically speed up debugging workflows. He’ll also share how a side project born from personal frustration - evolved 🚀 into a real AI-powered engineering system used at scale: "AI-Native Development: From Understanding Code to Fixing It Autonomously" with Guy Bary and @IsraelZablianov wix.engineering/ai-native-deve…

Most AI teams don’t have a model problem. They have a context engineering 🚩 problem. Bigger context windows don’t automatically mean better AI results. In fact, poorly structured context can increase costs while hurting accuracy. In this @Wix Engineering Conference 2026 session, Maya Halevy explains what really happens inside LLMs - from attention limitations to the "Lost in the Middle" problem - and shows how to structure prompts, instructions, examples, and retrieved data for better precision and lower cost. Watch her talk to learn practical techniques to: ✅Improve model accuracy ✅ Reduce token waste ✅ Structure context more effectively ✅ Build more reliable AI systems If you work with AI agents, RAG, copilots, or LLM-based products, this talk will change the way you think about prompt and context engineering:

.@SimonsDalia's session "The Pinsky Method: A Task Division Strategy to Minimize Context Switching and Maximize Developer Flow" was awesome. The video will be shared soon 😀

אני מראיינת השבוע את @aviranm על מודל הגילדות החדש של וויקס. מה תרצו שאשאל?

The industry assumes AI agents need "skills" to work well. So we at @Wix ran 250 evals to 🧪 test it. The surprising 👀 result? In many cases, optimized docs outperformed the skills themselves. Why? 🤔 Because skills age fast. A small mismatch between the skill and the actual product/API was enough to send agents into loops, wrong implementations, or unnecessary complexity. Even more interesting - skills sometimes made agents less effective - because they stopped exploring better solutions outside the prescribed path. The takeaway: Docs are the foundation. Skills are just an acceleration layer. In the AI era, documentation is no longer just for humans. It’s infrastructure for agents. Read this excellent study by @afriedmannsays: wix.engineering/post/we-ran-25…

Our meetup 👉 next week: AI is already writing code - but without the right context, it’s still guessing. Developers waste hours digging through codebases, and bugs take days or weeks to go from a user complaint to an actual fix. In this meetup, you’ll see what it looks like when that changes. Join @guy_bary and @IsraelZablianov to learn how to give AI real context so it can actually understand your code, explore entire codebases in seconds, and become a true research partner. And on the other side - what happens when bug fixing becomes almost autonomous, shrinking the path from issue to Pull Request dramatically. If you want to understand where development is heading (and stay ahead of it) - this is the meetup to be: wix.engineering/ai-native-deve…

What if intelligence was just… language? At the Wix Engineering Conference 2026, @GiladBarkan, Head of @Wix's Data Science Guild, took us on a "Tour de AI" - not another tools talk, but a deep dive into why AI works. The session connected disciplines - from NLP and linguistics, through math, philosophy, and economics - all the way to modern LLMs and how we, as developers, actually interact with them. Watch this insightful session to build a clear mental model of how language became the foundation of modern AI - and understand why LLMs work, and where they break. This wasn’t about using AI. It was about understanding it - so you can use it better. And it all led to one big question: Is language all you need for intelligence?

🚩המיטאפ שלנו בשבוע הבא: AI-Native Development: From Understanding Code to Fixing It Autonomously - עם @guy_bary ו- @IsraelZablianov בקמפוס וויקס (11.5) לפרטים והרשמה: wix.engineering/ai-native-deve…

AI, real production systems, and an engineering conference with 1,000+ engineers - here’s what’s inside our latest Wix Engineering newsletter (Edition 86): 🚀 Wix Engineering Conference 2026 1,000+ developers. One day. Pure engineering energy. A day that started with one question - how is AI reshaping engineering - and turned into a full deep dive into real production systems, built and powered by @Wix engineers. 🧠 AI-Native Development - our upcoming meetup AI writes code - but without context, it’s guessing. Join @guy_bary & @IsraelZablianov (May 11) to see how AI understands codebases and starts fixing bugs autonomously. ⚙️ Data-to-Production From Iceberg to live microservices: a 3-layer pipeline enabling sub-second queries on billions of rows - replacing weeks of glue code. By Almog Gelber 💡 insights we liked * AI Coding Agent Manifesto - verification and coding (Shay Cohen) * Why books still matter in the AI era (@NirOrman) * Working with AI - partial understanding (@schuftan ) * MCP Resources - data layers matter as much as actions (Elena Kravchenko) Read the full edition: wix.engineering/so/a5PtPxWGM