I build AI systems that take on everyday work in a business: bots that answer customers on WhatsApp and your website, agents that move information between your CRM, ERP and email, and search that finds answers in your own documents. Every system is tested on real conversations and real data before it meets a customer. 15 years in software, 8 of them leading teams, up to VP R&D.
AI news in Hebrew, by email every morning.
The important updates from the big AI companies, with a plain-language note on what each one changes for a business. Get it every morning, once a week or once a month.
Sign up for the brief
The open-source projects developers everywhere are starring right now, and the standouts in each field, each with a short summary.
A page per model, with what changed lately and what it means for a business, in plain words.
The freshest model announcementGeminiGoogle releases Gemini 4 CarbonWhat's new in Gemini →We start with the process that weighs on the business most today, prove it works, and expand from there.
For exampleA customer asks on WhatsApp at ten at night what it costs and when they can come in, and gets an exact answer from your price list without waiting for the morning.
For exampleA lead from the website lands in the CRM and gets a follow-up message, without anyone copying details between windows at the end of the day.
For exampleA new employee asks how to handle a refund and gets the answer from your own procedure, with a pointer to the section it came from.
For exampleOne system that holds customers, orders and stock exactly the way your business works, in place of several subscriptions that don't talk to each other.
For exampleCampaigns are checked every day, weak ads are paused in time, and you get a short summary in plain words.
For exampleYour team learns the tools on its own real tasks, with clear rules on what may go into a model and what stays in-house.
Every project goes through the same four stages. A system goes live only after it has been measured on real data.
We talk about the business: where the time goes, what customers ask again and again, and what information already exists.
A one-page drawing of the system: what goes in, what comes out, and how we'll know it works. Then a small pilot on real data.
We run the system on real past questions and measure whether the answer is right, how long it took, and where it gets stuck.
The system goes live with documentation, monitoring and one contact who knows it in depth. After launch I stay on for changes, improvements and questions.
After rev.C the stamp goes on: the system is tested and working, and someone is accountable for it.
Craft Anything: understand how any software works, and craft your own.
The decisions behind the site: App Router, static pages, two languages, SEO, and keeping it fast as the content grows.
All breakdowns →Send a question about AI in your business and the answer runs in the brief. With your name or without it.
Send a question →I've been writing software for 15 years, 8 of them leading development teams up to VP R&D. These days I build AI systems for businesses, from customer service bots to agents connected to the CRM and ERP.
In every project the goal is that the team stops answering the same questions and copying the same data, because a system does it correctly and on time. That is why every project starts with a drawing and ends only after it has been measured on real data.
Away from the keyboard I play piano and guitar, and volunteer with the Border Police.
If you want the details anyway: CV
Tell me about it in a short call. We'll sketch together how it could work.