מאמרים מאת Guy
פעילות
9K עוקבים
-
Guy Korland שיתף את זהCongratulations! You’ve all just been promoted! The era of the "Solo Dev" is quietly evolving into the era of the AI Team Lead. I’ve been watching how the most successful developers interact with agents like Claude Code, Cursor, Copilot, or specialized orchestrators. There is a distinct "click" that happens when a developer stops writing line-by-line and starts acting like a manager. If you want to stay productive in 2026, you need to adopt these three leadership principles: 1. Kill the Serial Workflow The biggest mistake you can make as a new "manager" is working in a serial way. In the old days, you did Task A, then Task B. If you do that now, your "employees" are sitting idle while you're busy thinking. * The Shift: You should have 3–5 agents running in parallel across different branches or Git worktrees. * The Goal: While Agent A is refactoring the legacy API and Agent B is generating unit tests, you should be at the whiteboard (or the architecture doc) planning the next sprint. Never let your compute power sit idle. 2. Respect the Span of Control In the physical world, the "sweet spot" for a manager is 6–8 direct reports. Try to manage 20, and you become the bottleneck. AI agents are no different. * The Trap: Trying to juggle too many agentic workflows at once leads to "contextual collapse." * The Strategy: Just like human devs, you won't follow them live or "shoulder surf" while they code. Give them the autonomy to execute, just ensure they provide a comprehensive "summary of work" and a clean PR at the end. 3. Manage Someone "Smarter" Than You This is the hardest pill for most of us to swallow. In terms of raw execution speed and syntax memory, your "employees" are now faster and more knowledgeable than you. * What they DON'T need: Empathy, small talk, or a sense of personal accomplishment. * What they DO need: High-level leadership, creative innovation, and the specific "North Star" direction that triggers their ability to execute. The Future of CS Education? We still need undergraduate students to learn the fundamentals of Computer Science, you can't manage a team if you don't speak the language. However, the curriculum must pivot. Software Engineering and Project Management are no longer "soft skills", they are the core technical requirements for the modern developer. The prompt is no longer just code, it’s a leadership directive. Are you still writing, or are you leading?
-
Guy Korland פרסם/פרסמה את זה מחדשGuy Korland פרסם/פרסמה את זה מחדשWOW! Thanks Gartner for recognizing FalkorDB as a leading GraphRAG vendor! The newest Gartner report on #GraphRAG has a clear message: if you’re building AI applications, combining Knowledge Graphs with LLMs is today’s smartest strategy. Key insights: - GraphRAG enhances retrieval performance by 35–46%, especially for complex, multi-hop, and temporal queries. - Integrating Knowledge Graphs with LLMs not only boosts accuracy but also reduces hallucinations and enables deeper, more reliable reasoning. - Organizations not adopting graph approaches will carry ± 75% more AI technical debt than those sticking to conventional methods.
-
Guy Korland פרסם/פרסמה את זה מחדשGuy Korland פרסם/פרסמה את זה מחדשChecking out Mem0 for your memory layer? The mem0-falkordb plugin provides: 1. Graph-structured memory: Store relationships between entities, not just flat facts 2. Per-user graph isolation: Each user gets their own isolated FalkorDB graph 3. Context-aware retrieval: Semantic search with vector embeddings 4. Memory evolution: Support for updates and conflict resolution 5. Runtime patching: No modifications to Mem0 source code required Guy Lubovitch presents all of this (and more) at AI Agents SF, 20 April! (Thanks Bay Area AI!). Link in the comments.
-
Guy Korland שיתף את זהפרויקט סופ"ש: האם בינה מלאכותית באמת מאיצה פיתוח קוד פתוח? תשארו איתי – קצת "חפירה" הכרחית על המצב, ואז אשתף מה בניתי. לאחר שאנטרופיק הציגו את המודל החדש Mythos (הם טוענים שזה המודל הכי מדהים ומפחיד שהם ראו, אבל שיווק זה תמיד חלק מהמשחק), הם הראו כיצד הוא הצליח למצוא באגים וחורי אבטחה קריטיים בפרויקטי קוד פתוח שרצים עשרות שנים. אלו באגים שהסתתרו שם זמן רב, ומי יודע אילו גופי ביטחון או פושעי סייבר ניצלו אותם עד היום. ההכרזה הזו נגעה בדיוק בשאלה שמעסיקה אותי כבר זמן מה: האם ה-AI כבר הגיע לשלב שבו הוא באמת מאיץ פיתוח של פרויקטי תשתית מורכבים עם קוד גדול וותיק? ב-FalkorDB אנחנו כבר מרגישים את ההשפעה בכל פינה. אנחנו מייצרים כלים ומוצרים חדשים בקצב כמעט לא הגיוני, שלפעמים אפילו מאתגר את המיקוד של החברה (אבל זה נושא לפוסט אחר). עם זאת, כשזה הגיע לקוד הליבה שלנו, קוד ב-C, יעיל ואופטימלי עד הקצה לא ראינו עד לאחרונה השפעה רחבה. בכל פעם שצץ באג חדש ניסיתי "אייג'נט" אחר, ורוב הזמן הם חזרו מהר מאוד, מרוצים מעצמם, עם פתרון "פלסטר" שרק מכסה על הסימפטום בלי להבין את שורש הבעיה. אבל מאז ש-Opus 4.6 יצא, התחלנו להרגיש שמשהו השתנה. פתאום הוא הצליח לנתח באגים קשים שבאופן רגיל דורשים שבועות של עבודה, באגים שדורשים הבנה עמוקה של קוד מקבילי, סנכרון ואינטראקציה לפעמים בין פרויקטים שונים. זו בדיוק ההאצה שחיכינו לה. אבל יש עוד צד למטבע שיכול לפגוע בקהילות הקוד הפתוח: Slop PRs. אנחנו מקבלים כמות הולכת וגדלה של "תרומות קוד" שברור שנוצרו במלואן ע"י AI, מבלי שהתורם טרח לבדוק אם הן בכלל פותרות את שורש הבעיה. הנזק כאן כפול: בזבוז זמן יקר של המתחזקים (Maintainers) שנאלצים לעבור עליהן ולפסול אותן. שחיקה ואדישות מצד המתחזקים, שגורמת להם לעיתים להתעלם גם מתרומות איכותיות ומשמעותיות, מה שיוצר תסכול בקהילה. זה נושא שחייב להטריד את עולם התוכנה. בסוף, רוב התוכנה בעולם נשענת על פרויקטי קוד פתוח. אם הם ידעכו או יינטשו בגלל העומס הזה, כולנו נרגיש את ההשפעה. אז אחרי כל הדיבורים, התיישבתי בשישי האחרון יחד עם קלוד ותכננו פרויקט שינסה למדוד את ההשפעה האמיתית של סוכני AI על פרויקטי קוד פתוח (בדגש על פרויקטי תשתית ותיקים). ביקשתי ממנו לאסוף מידע מ-GitHub כדי לייצר גרפים שיציגו את המגמה אל מול התקדמות ה-AI בשנים האחרונות. החלטנו למדוד מספר מדדים מרכזיים: יחס כמות הקוד למפתח: בהנחה שבינה מלאכותית אמורה להגדיל את התפוקה. יחס PRs שמוזגו (Merged) למפתח: כדי לוודא שיש ערך אמיתי ולא רק "רעש" של ייצור שורות קוד. קצב סגירת PRs ללא מיזוג: מדד לרמת ה-PR Slop. כמות התורמים לפרויקט: כדי לראות אם ה-AI מעודד או פוגע בצמיחת הקהילה. אחרי יומיים שבהם קלוד כתב קוד ואני בעיקר לא הפרעתי לו, נולד AgentPrint. כמובן שהדבר הראשון שעשינו היה למדוד את ההשפעה על FalkorDB, והמספרים מאשרים את מה שהרגשנו: כמות הקוד למפתח עלתה, כמות ה-PRs לפתח עלתה, אבל גם כמות ה-Slop PRs שאנחנו דוחים עלתה משמעותית. אני מאמין שנראה שיפור באיכות ככל שמודלים כמו Opus 4.6 יכנסו לשימוש רחב יותר. תוצאות לייב וקישור לדוחות של פרויקטים נוספים כגון Redis, Ollama מחכים לכם בתגובות!
-
Guy Korland פרסם/פרסמה את זה מחדשGuy Korland פרסם/פרסמה את זה מחדש29 APRIL: Learn how to combine CocoIndex's incremental data transformation capabilities with FalkorDB's graph and vector search to build scalable, low-latency GraphRAG systems that keep AI models grounded in fresh, structured knowledge. - Q&A - For engineers who need to keep knowledge graphs in sync with changing source data - For anyone interested in incremental data pipelines for AI Link in the comments, see you there!
-
Guy Korland שיתף את זהAgents are doing the work, but humans still need the insight. 🧠🤖 The industry is currently obsessed with "Agentic workflows." We are building systems where agents talk to agents, solve problems, and navigate data autonomously. It’s an incredible shift. But at FalkorDB, we believe in a crucial principle: The more we rely on AI agents, the more important the Human-in-the-Loop becomes. "Human-in-the-loop" shouldn't just be a safety fallback, it should be an empowering experience. If you can’t easily visualize, query, and understand what your agents are doing inside your knowledge graph, you’re flying blind. That’s why we are continuing to invest heavily in the FalkorDB Browser. Our April 2026 updates are officially live, focusing on delivering an amazing UX that brings "explainability" to life. We’ve streamlined how you interact with your data so that even as your AI ecosystem grows in complexity, your personal understanding of the graph remains crystal clear. What’s new in this update? ✅ Enhanced visualization for complex pathfinding ✅ Streamlined UI for faster manual exploration ✅ Improved workspace management for agent-generated insights In an automated world, clarity is a superpower. Check out the full update notes here: https://lnkd.in/dWS_uHRf #KnowledgeGraph #AI #GenerativeAI #GraphDatabases #DataVisualizationFalkorDB Browser Update – April 2026 Release NotesFalkorDB Browser Update – April 2026 Release Notes
-
Guy Korland שיתף את זהIs a "Personal Wiki" just a Knowledge Graph in disguise? 🤔 Andrej Karpathy’s idea of flat .md files "Second Brain" is incredibly popular right now. But we need to talk about the "Efficiency Gap." Karpathy’s solution: ✅ Readable ✅ Simple to start 📛 Inefficient at 10k+ nodes 📛 Massive token waste 📛 High latency for relationship-heavy queries If you have nodes and you have links, you have a Graph. Using an Agent to crawl flat files is a "brute force" approach to RAG. By moving to a GraphRAG architecture, you stop using tokens to "discover" relationships and start using them to "reason" over them. Elegance is great for a weekend project. Efficiency is required for a production-grade system. #AI #GraphRAG #Productivity #KnowledgeGraph
-
Guy Korland שיתף את זהHad a fantastic time yesterday at the AI-ready Data Summit! It’s always a pleasure to step out of the dev lab and discuss the successes (and the messy failures) of building AI-ready data layers with fellow builders like Paul Dix and Shirshanka Das. Thanks lakeFS for hosting such a high-energy event!Guy Korland שיתף את זהMeet the third panelist joining us at the AI-Ready Data Summit 👇 Guy Korland, PhD, is the CEO of FalkorDB, bringing 20+ years of experience in the database industry to the conversation. Before FalkorDB, Guy served as SVP and CTO of Incubation at Redis, where he developed deep expertise in distributed systems and AI infrastructure. He holds a Ph.D. in Computer Science from Tel Aviv University. On March 31st, Guy joins Paul Dix (CTO, InfluxDB) and Shirshanka Das (CTO, DataHub) for a candid panel on the real challenges, failures, and successes of building AI-ready data infrastructure — moderated by IT journalist Paul Gillin. 🗓 March 31, 2026 | 10 AM – 2 PM EDT 🎟️ Free virtual event: https://hubs.la/Q047YrSk0 4 hours. 10 sessions. 12+ speakers. Register free today. #AIReadyDataSummit #DataInfrastructure #FalkorDB #DataEngineering
-
Guy Korland פרסם/פרסמה את זה מחדשGuy Korland פרסם/פרסמה את זה מחדשJust got back from RSA Conference 2026 — what an experience. A few takeaways: 🔒 The cyber community is as vibrant as ever. The energy on the floor was incredible — so many passionate people working on hard problems. 🤖 AI is everywhere in conversation, but the actual adoption in security products still feels early. Which to me means one thing: the most interesting developments are still ahead of us this year. 🕸️ Graph is gaining real traction in cybersecurity. Many companies are already using it or at least aware of it — but we've barely scratched the surface of what's possible. Threat detection, identity graphs, attack path analysis... the use cases are massive. My prediction: by RSA 2027, we'll see both significantly more AI-powered security tooling AND a real shift away from SQL/NoSQL toward graph and LLM-native data architectures to tackle cyber challenges. Excited to be building at FalkorDB right at the intersection of all of this. 🚀 #RSA2026 #Cybersecurity #GraphDatabase #AI #FalkorDB
Guy Korland השיב לתגובה
14 H
Sergey Inozemtsev we find some of the AI code review tools super effective, and always make sure a couple of different tools review each PR
Guy Korland השיב לתגובה
23 H
Volodymyr Brodskyi I like the analogy and I can deeply connect on the personal level :)
But I think it’s less like kids and more like having five Green Lanterns all running in parallel, each with imagination-based constructs that can spin up anything instantly.
The upside: insane speed and creativity.The downside: if you don’t give them clear constraints and guardrails, they’ll happily “optimize” the house into something unrecognizable 😅
So yeah… maybe not fire suppression, but definitely containment and control are part of the new CS curriculum.
Guy Korland השיב לתגובה
1 ימים
Leonid Breido I don't think it's an holy grail, it's like hiring a super smart developer that still needs your leading, it will create bugs, or try to work around things from time to time, but won't human try to do it too from time to time?
-
Guy Korland אהבתי את זהGuy Korland אהבתי את זהFinally! A text-to-SQL solution that actually works! (100% open-source) When text-to-SQL fails, the real issue isn't the LLM or the prompt but schema retrieval. Consider a query like "Which publishers received royalty payments above $5,000?" To handle this, vector search can pull "publisher" and "royalty_ledger" based on semantic similarity But it can completely miss "vendor_agreement", the bridge table that connects them. The LLM writes valid SQL, but the engine still returns zero rows. This is a fundamental issue with vector-based schema retrieval on enterprise databases. A smarter approach is to treat the schema as a graph instead of a document to embed. Tables become nodes, foreign keys become edges, and join paths are discovered by walking the graph rather than matching semantics. If you want to see this in practice, QueryWeaver implements this approach. It converts the schema into a graph, and when a query comes in, it walks the structure and pulls in every bridge table the join path requires, including multi-hop chains. For instance, on the BIRD Benchmark with a superhero database expanded to 60 tables, it resolved a 5-hop query by chaining through: superpower → capability_matrix → stakeholder_registry → resource_requisition → budget_allocation Vector search found the two endpoints but missed everything in between because "stakeholder" has zero semantic link to "superpowers." Graph traversal found "stakeholder_registry" simply because it was the only road connecting the entities. It's fully open-source, and you can easily self-host it. I've shared the GitHub repo in the replies. ____ Share this with your network if you found this insightful ♻️ Find me → Avi Chawla. Every day, I share tutorials and insights on DS, ML, LLMs, and RAGs.
-
Guy Korland אהבתי את זהGuy Korland אהבתי את זה🏁 Just finished a community MCP server for Lettria's Perseus tool! This covers 29 different tools to interface with Perseus for things like: - Quickly building a graph from arbitrary text sources - Optionally using ontology .ttl files to build those graphs - Interlinking different graphs - Push the data to a Neo4j or FalkorDB database #mcp #knowledgegraphs #contextgraphs #ontologies
ניסיון חינוך
-
FalkorDB
צפו Guy בחוויה המלאה
ראה את כותרת, קְבִיעוּת ועוד.
או
ניסיון בהתנדבות
-
Member Board of Directors
Ezer Lahaim
- 3 שנים 7 חודשים
בריאות
Supported Ezer Lachaim digital transformation, moved from manual call center to mobile Transportation app.
פרסומים
-
Practical Solutions for Multicore Programming From Crafting a High-Performance Ready-to-Go STM to non-Linearizable Data Structures
Guy Korland
ראו פרסוםThis dissertation presents several novel techniques for practical multicore programming in a shared memory multicore machine.
The basis of the research initiative, which is covered in the first part of this dissertation, is Deuce [50], a novel open-source Java framework for Software Transactional Memory (STM). Deuce has several desired features not found in earlier Java STM frameworks. When this research was started there did not exist an efficient full-featured Java STM framework that could…This dissertation presents several novel techniques for practical multicore programming in a shared memory multicore machine.
The basis of the research initiative, which is covered in the first part of this dissertation, is Deuce [50], a novel open-source Java framework for Software Transactional Memory (STM). Deuce has several desired features not found in earlier Java STM frameworks. When this research was started there did not exist an efficient full-featured Java STM framework that could be added to an existing application without changes to its compiler or libraries. It seemed that any efficient Java STM would require compiler support. -
Deuce: Noninvasive Software Transactional Memory in Java
Transactions on HiPEAC
שפות
-
Hebrew
-
-
English
-
המלצות התקבל
1 אדם המליץ Guy
הצטרפו עכשיו כדי נוף