VP R&D | Agentic AI products, built by agentic AI teams | Data eng. I ❤️ #AI, #BigData #DataWarehouse’s #DataLake’s and all in between. Amateur #astronomer 🪐🔭

Joined April 2009
Cool tip!
𝗖𝗹𝗮𝘂𝗱𝗲 𝗖𝗼𝗱𝗲 𝘁𝗶𝗽: once Opus 5.5 is your main model, don't let your Fable quota sit idle. Put Fable on call with /advisor. Run /advisor fable and Claude Code gets 𝗮𝗻 𝗮𝗱𝘃𝗶𝘀𝗼𝗿 𝗼𝗻 𝘁𝗵𝗲 𝘀𝗶𝗱𝗲𝗹𝗶𝗻𝗲𝘀. Opus 5.5 keeps doing the work and usually calls Fable in at three moments: → Before committing to a plan: is it right? → When the same error keeps coming back: am I heading the wrong way? → Before saying "done": did I miss anything? Fable reads the whole session and gives advice. Opus still writes the code. The tree is the full setup. Opus 5.5 on high runs the main session. Three subagents on medium split the rest: one reads code, one edits and runs tests, one looks up docs. Fable waits on the sidelines. Hand the tree and this prompt to Claude Code to set it up 👇 "Set up my Claude Code to match this tree: 1. For the three subagents, reuse fitting ones from ~/.claude/agents and .claude/agents, and propose new ones only for missing roles. Set each to model: opus and effort: medium. Leave any subagent that already sets a different model alone, and list it. 2. Set the main session's default effort to high with effortLevel in ~/.claude/settings.json. 3. Check for environment variables that turn the advisor off (CLAUDE_CODE_DISABLE_ADVISOR_TOOL, or DISABLE_TELEMETRY and anything else that stops flag fetching) and for CLAUDE_CODE_EFFORT_LEVEL, which overrides every subagent's effort. Report them; don't change them. 4. Add one rule to ~/.claude/CLAUDE.md: ask the advisor before committing to a big plan, when the same error shows up a second time, and before calling a long task done. Show me the changes first. Don't edit any files yet." code.claude.com/docs/en/advi…
7
Yaar Reuveni retweeted
Holy crap. Rick and Morty just explained Jev AI to me better than any tech demo could.
254
1,316
284
11,432
1,639,194
True story. And I would add throw it on Claude is already the 2025 realization today you should be having a team of agents building for you autonomously in the background while you are surfing your Claude on the tough tasks
Your job may be surviving on inertia. Your resentment won’t extend the runway. What do you actually want to do with the intelligence now at your disposal?
1
41
Yaar Reuveni retweeted
I'm Boris and I created Claude Code. Lots of people have asked how I use Claude Code, so I wanted to show off my setup a bit. My setup might be surprisingly vanilla! Claude Code works great out of the box, so I personally don't customize it much. There is no one correct way to use Claude Code: we intentionally build it in a way that you can use it, customize it, and hack it however you like. Each person on the Claude Code team uses it very differently. So, here goes.
1,306
6,929
1,280
54,202
8,236,304
Wow
Today, we're announcing four new capabilities on the Anthropic API to help developers build more powerful AI agents. A code execution tool, MCP connector, Files API, and extended prompt caching:
62
This is wild! 😲
Introducing Claude 3.7 Sonnet: our most intelligent model to date. It's a hybrid reasoning model, producing near-instant responses or extended, step-by-step thinking. One model, two ways to think. We’re also releasing an agentic coding tool: Claude Code.
36
Yaar Reuveni retweeted
1 בשבוע שעבר נפגשנו, קהילת הופכות למפתחות, למפגש בנושא How to Kickstart Your Career. המפגש הועבר על ידי סיגל ויינר (מעון), מנהלת פיתוח בכירה ב @awscloud ואחת ממנהלות קהילת @Baot_IL , ונערך במשרדי חברת @EleosHealth - החברה בה אני עובדת 🙂 >>
1
1
1
13
495
Earlier this week our Amit Spinrad gave an amazing (Hebrew) talk in a Meetup part of MeDS - Medical Data Science Israel community - highly insightful talk about our #ml and about how you can use specialized generative #ai models and #llms in real life #be…lnkd.in/d5MqGEdH
51
We are expanding our DevOps team - looking for a senior DevOps preferably with MLOps experience. Know anyone who might be interested? Also, reminder our Data engineer position is still open ;) lnkd.in/datFymCh #hiring #devops #mlops #generativeai…lnkd.in/dMZAG7Qe
38
Is #chatgpt or #googlebard good for everything? Maybe😉 But sometimes smaller and cheaper models perform great for specific use cases, As our Amit Spinrad quoted in his new blog post “Jack of all trades master of none“ - inviting y…lnkd.in/dhAv3CPx lnkd.in/dcPr4htt
1
62
Yaar Reuveni retweeted
How to present In 2006, I helped @ericschmidt create a deck outlining Google’s strategy, for a presentation Eric was delivering to the company. It taught me a profound lesson on how to present. When I showed up to my first meeting with Eric, he asked me to visit with every product team at Google, chat with them to figure out what they were working on, and then summarize it on one slide (for each team). Easy enough, I thought. I would use 3-5 bullet points per slide. Piece of cake. I started mentally mapping things out and got ready to leave. “But”, Eric said, “I want no words on any slide”. My well-laid plans disintegrated in an instant. How was I supposed to convey the key messages from each team, without WORDS? Eric must have seen the panic on my face, and kindly gave me a hint. “Put the text in speaker notes”. “But what goes on the slides, Eric?” I continued panicking. That classic, gentle “Eric smile” fluttered on his face. “Why, images, of course!” “You mean, you want each slide to just be comprised of images?” “You got it. And use the title wisely. 7-8 words max. Let’s meet in a week to review progress.” As I left the meeting, little was I to know that this conversation would fundamentally change my view on how to deliver effective presentations. 17 years later, I still cling tightly to the following principles: 1. The larger the audience, the fewer the words on the slide. In Eric’s case, the audience was thousands of employees, so we had 0 words per slide. 2. The title does most of the heavy lifting, which means it cannot be passive. It must be action oriented. Eg: not “Subscriber retention” but “Subscribers continue to be retained strongly” or even better “Net revenue retention continues to be > 100%”. 3. Use memorable images that substantiate and give credence to the words of the title. This image is what will occupy most of the slide area, so you need to spend much of your time thinking about what picture will best get the point (made by the title) across. In some cases, it might be a customer image or logo. in other cases, a graph. In yet other cases, it could be something else entirely. For the Google presentation, one of the images that gave me the most trouble was a slide on Google Search Appliance and other Enterprise products. The title stated that these products were increasingly being used by larger customers. The team didn’t want to share customer logos broadly since some were confidential, so logos were not an option. I decided to go with a trend line on the % of searches from enterprise customers, but the person who was supposed to pull this data for me, flaked at the last minute and I had to scramble. I ended up scrambling to create a mosaic of a bunch of consumer product logos with some kind of icon that denoted large enterprises. Not my finest moment but it got the point across. 4. Use speaker notes. Like Eric said, speaker notes should contain most of the details. It puts a lot of burden on the speaker since they cannot just read off the slides. But this doesn’t deter good speakers, since they prepare dozens of times, and then again. So there you have it: my 4 principles for delivering compelling presentations to live audiences. (CAVEAT: If the presentation has to be emailed to an audience who will consume it asynchronously, that’s completely different and has different rules). How did the 2006 Google strategy presentation turn out, you ask? It went quite well, and later I got a nice thank you note from Eric. I didn’t realize at the time that I should have been the one thanking him for the once-in-a-lifetime learning opportunity.
133
418
82
3,292
892,327
ה chief clinical officer התותחית שלנו @ShiriSharvit התארחה בחדשות כאן לשתף על הקשר בין צ׳אטבוטים וטיפול נפשי! מאד מעניין! #ai #chatbot #chatgpt #behavioralhealth
תמיד שם בשבילך: האם בינה מלאכותית יכולה גם להעניק סיוע נפשי? הפסיכולוגית שירי שדה שרביט הסבירה ל-@Yoav__Zehavi על היתרונות והחסרונות ב-#העולם_היום
1
82
(1/4) #BookRecommendation: Recently finished reading this excellent tech #leadership book a dear friend of mine (@ItaiYaffe ) brought me as a gift from one of his conference trips, the book is “Debugging Teams” by @therealfitz and @sussman
1
3
(3/4) As a taste, I used to grow sourdough from an excellent starter I hunted from a good friend who had been growing his for years, I especially related with the analogy @sussman and @therealfitz made about how growing a great team is like growing sourdough -
1
(4/4) you need to begin with a great starter and then consistently nourish and care for it, every new flour you add will absorb what you have already built, the bacteria culture you had in your starter will overtake and you can grow more excellent sourdough and bake amazing bread