Wix Engineering

Wix Engineering's official Twitter. We develop innovative cloud-based web applications that influence our 250M+ users worldwide

[Avishai Abrahami @Avishai_ab]

We are bringing @Wix's powerful tools into more hands as a SpaceX Super Intelligence Launch Partner.

As a @spacexai partner, we're bringing Wix's capabilities, including Wix Headless, into agentic workflows. In @grok, that means making it easier to build, customize, and manage web experiences with Wix using natural language

[Nir Zohar @nirzo]

We moved into our new @Wix New York office today. It looks great. But honestly, it's the same team that was in the old one, and that's what makes it feel like home. Thanks to everyone who pulled this off.

[Wix Engineering @WixEng]

Every hour spent managing website security is an hour taken away from growing a business. At @Wix Engineering, we build security into the platform - from 24/7 threat monitoring to automatic patches and DDoS protection - so you can focus on your growth.

Our VP Engineering, @aviranm on why built-in security is a strategic advantage:

[Aviran Mordo @aviranm]

Wix’s 24/7 security team continuously monitors, strengthens, and maintains the platform, so you never have to worry about keeping your site secure.

[Aviran Mordo @aviranm]

Wix’s 24/7 security team continuously monitors, strengthens, and maintains the platform, so you never have to worry about keeping your site secure.

[Aviran Mordo @aviranm]

If your website is running on @Wix , you are safe and not effected by this vulnerability.

[Wix Engineering @WixEng]

What does it take to build agentic products on open-weight models?

@ittaiz is speaking at @NVIDIA GTC Berlin 2026! 👏

He'll join a panel exploring why builders are moving to open-weight models, the trade-offs in cost, latency, and quality, and what it takes to run them safely at scale - including what stays on closed models and why. Good luck, Ittai!

[Wix Engineering @WixEng]

We don't. Many filter conditions are a simple yes/no in code but part of the skill’s internal logic, so they're hard to phrase well for the LLM. And in practice it didn't need them: with the KB and user context, it answered based on what the next step of the conversation required.

[Wix Engineering @WixEng]

An AI skill can match a request perfectly - and still be impossible to execute. @AshkenaziOrtal’s new @arxiv paper explains how @Wix's Helpmate checks account state before skills reach the LLM.

Across 756K+ production messages, this filtered out 59.4% of matched candidates and reduced skill context by 90.5%, saving 228.8M tokens.

At @Reversim Summit 2026, Ortal will explore this approach as part of her talk on designing, routing, and scaling AI skills in production.

Read the papert: https://t.co/f7galp81D7

[Wix Engineering @WixEng]

From building production AI agents to hiring the engineers who will build what comes next >>> we're proud of @AshkenaziOrtal and Guy Nahum, who are bringing their experience to @Reversim Summit 2026! 👏 👏

🎤 Ortal will explore what happens when one LLM skill becomes hundreds: how to define skills, catch common failures, and route requests to the right capability in a system handling 300K conversations a month. Her talk builds on research co-authored with Vitalii Kloz and Mykhailo Ulianchenko, published on @arXiv.

🎤 Guy will tackle "Hiring Post-Claude": what makes a great developer today, why junior engineers matter, and how @Wix is rethinking hiring and development programs to cultivate the #xEngineer.

Heading to #Reversim? Don't miss their sessions!

[Wix Engineering @WixEng]

Building with AI is getting easier. Understanding why it worked - and making it work again - is the harder part.

On Nov 2, Amir Harel (Symphony by @Wix) will take the stage at the Claude Community meetup at Wix Campus, TLV, to show how a real feature evolved from a successful experiment into a repeatable process.

Good luck, Amir! Details & RSVP:
https://t.co/YEIhZKaYOT

[Wix Engineering @WixEng]

איך חיזוי המילה הבאה הפך לבינה מלאכותית שיודעת לכתוב קוד ולהפעיל תוכנות?

גלעד ברקן ראש גילדת הדאטה סיינס ב-Wix הנגיש את המאמר החדש שלו - From Autocomplete to Intelligence: How Language Became the Operating System of AI - למסמך פתוח בעברית, בשביל שכולם יוכלו ללמוד.

המאמר של גלעד מסביר איך מודלי שפה לומדים, למה השפה נמצאת בלב מהפכת ה-AI, ואיפה היכולות שלהם עדיין מוגבלות. לא צריך ידע מוקדם, רק סקרנות. מוזמנים לקרוא, לשאול ולהשאיר הערות במסמך.

https://t.co/T42sf6Z8B1

[Wix Engineering @WixEng]

How do you move critical 🚩 production databases between clusters while millions of users keep using the applications that depend on them?

At @Wix, we needed to solve this to improve reliability: when applications share a database cluster, a problem in one can affect the others.

Bar Shauli shares how she and our Wix Data Infrastructure team built an internal migration service using Kafka and ProxySQL on top of MySQL, allowing databases to move safely between clusters with zero downtime.

From streaming data and handling partitioning challenges to switching live traffic, here's how we make database migrations happen at massive scale-without users noticing.

👉 Read her great blog post:
https://t.co/8zl5Z7N4em

[Yoav Abrahami @yoavabrahami]

Components are dead!!!

Agents work better with something completely different - UI Regions.

Read all about the new pattern:
https://t.co/voV2z8ouDa

[Wix Engineering @WixEng]

Components help developers manage complexity. But what happens when AI agents have to navigate the boundaries we created?

Scattered files, slots, and context can force agents to reconstruct a page before they can safely change it. In this article, @yoavabrahami explores how Jay, a new frontend framework, approaches that challenge through regions: keeping markup local and editable while sharing the underlying logic.

You’ll learn how template flattening moves composition from runtime to coding time, how targeted updates keep pages reactive, and how validation and sync preserve local edits as templates evolve. Read here:

https://t.co/HiaiM7xdBx

[Wix Engineering @WixEng]

What does it take to build agentic products on open-weight models?

@ittaiz is speaking at @NVIDIA GTC Berlin 2026! 👏

He'll join a panel exploring why builders are moving to open-weight models, the trade-offs in cost, latency, and quality, and what it takes to run them safely at scale - including what stays on closed models and why. Good luck, Ittai!

[Ran Bar-Zik @barzik]

זה קצר וממש מעניין לקריאה ועושה סדר. אז בהחלט מומלץ לקרוא.

[Wix Engineering @WixEng]

What does it take to make on-call better - and turn production incidents into lessons that stick?

In "How to Win in Production" Guy Nahum walks through real incidents we've faced at @Wix, the failure patterns they exposed, and the changes that helped us reduce noise and learn without blame.

Watch this talk for practical rules and habits you can bring to your team, plus a grounded look at where AI can responsibly help:

https://t.co/exW1GD20N5

[Nir Zohar @nirzo]

I like getting shit done. So do our agents.
@Symphony_co_ by @Wix now connects to @link,
@stripe's wallet for agents.
That means our agents can finish the job:
reorder stock before it runs out, pay a supplier invoice the day it lands, cover an event entry fee before the deadline.
You approve each payment from your phone. The agents never see your card, and every transaction is capped at the amount you approved.
https://t.co/9v13oJyd9t

[Guy Bary @guy_bary]

I'm doing a major refactor of Octocode, and I'm trying a new approach to check tool schemas and descriptions, which works really good for me.

𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐈𝐧𝐭𝐞𝐫𝐯𝐢𝐞𝐰𝐬

I use large models (Claude Opus, OpenAI Astra,..) to create tools protocols.

Then I give those specs to smaller models (Haiku, Luna…) and interview them.

I force reasoned answers to:

- What did you understand?
- What's unclear or ambiguous?
- What assumptions could break?
- Where is it incomplete?
- What edge cases will you fail on?
- When would you ask the user for clarification?
- How do you measure success/failure?
- Can you generate outputs that strictly follow the protocol?
- Can you execute it?

Their answers expose gaps the big models missed.
Then I fix the gaps and repeat

Cleaner way to improve complex agent behavior 🐙

P.S - I will ship this methodology to the next prompts optimizer skill

https://t.co/9UGQoVKGyH

[Avishai Abrahami @Avishai_ab]

Nobody starts a business because they love paying supplier invoices on Friday afternoons.

So we're taking that off your plate. @Symphony_co_ by @Wix now connects to @link's wallet for agents by @stripe, our agents handle the payment, you approve it in seconds, and it never touches your to-do list again.

Build the thing you actually started this for. We've got the rest.

https://t.co/gkVv483UMO

[Wix Engineering @WixEng]

Proud to see @gilecks sharing his expertise at @ReactAdvanced London 2026! 👏

Gil will explore how React 19s use API can help improve application performance by controlling when resources load, components render, and pages become interactive. Don't miss his session!

https://t.co/pWMAesnaCa