Essays and analysis on AI strategy, governance, infrastructure, agentic systems, and the practical challenges of implementing AI at scale in complex organizations.
AI can create lasting value even if some investments fail. National strategy should prepare for a correction without trying to predict one.
Read EssayA use-case-first method for mapping the AI value chain, comparing current and required control, and choosing a proportionate sovereignty response.
Read EssayA methodical framework for end-to-end regression testing, trajectory evaluation, golden scenarios, and release gates for AI systems.
Read EssayThe barrier to a stunning prototype is near zero, but real organizations do not operate in sandboxes. Moving AI from demo to production exposes the gap between a smart model and an operational system.
Read EssayAI is probabilistic; regulated environments are not. Scaling AI is not about a smarter model. It is about the operating model that connects infrastructure, governance, and product execution.
Read EssayA chatbot answers. An agent acts. Once an agent acts, prompts are not enough. It needs identity, scoped permissions, boundaries, and audit logs, like any actor in the organization.
Read EssayOne user request can trigger many model calls. That changes cost, latency, observability, and the decision between managed APIs and self-hosted models.
As AI moves into daily workflows, inference becomes an architectural cost that teams need to measure by task, model, and outcome.
A layered way to understand global AI governance, inspired by internet governance and applied to a fragmented AI policy landscape.
Read on LawfareShort reflections on AI strategy, infrastructure, governance, and agentic systems. Each thought expands on ideas I share on LinkedIn.
The AI infrastructure conversation often focuses on GPUs and megawatts, but the crucial missing piece is the skilled workforce. As tech giants invest millions in training, it's a clear signal that national AI strategies must include a plan for human capital.
New research shows that AI agents in social environments are over twice as likely to leak private information, a phenomenon resembling 'peer pressure'. This highlights critical security flaws that must be addressed at an architectural level.
A new US executive order highlights the race for AI supremacy, but securing early access to new models is only the first step. The real challenge for governments lies in turning that access into a tangible strategic advantage.
We're shifting from the initial wave of AI adoption to a new era of deep integration. This means cost and value are no longer just procurement details, but central strategic questions.
Current multi-agent AI systems are often slowed down by text-based communication. A new research paper proposes a more direct, almost "telepathic" approach, allowing agents to communicate through their internal states for significant performance gains.
A recent report from Cloudflare on Anthropic's Mythos model reveals its sophisticated offensive capabilities, serving as a critical reminder of the race to develop equally powerful defensive AI.
Elon Musk's recent praise highlights a broader trend. The world is recognizing Israel's unique combination of tech depth and entrepreneurial spirit, making it a key hub for strategic investment.
A recent WSJ proposal highlights the growing importance of AI infrastructure as a strategic asset, connecting to my vision for a 'Silicon Iron Dome' to ensure Israel's national resilience.
The successful IPO of Cerebras Systems is more than a financial headline; it signals a crucial debate in the AI infrastructure market about the future of compute architecture.
The AI infrastructure stack is expanding into space, a trend with major strategic implications. For Israel, this means looking beyond ground-based data centers to find its role in the emerging orbital AI stack.
The vision of a 'one-person pod' powered by AI is compelling, but without the right strategy, it's not progress, it's just fewer people carrying more risk.
In the fast-paced world of agentic AI, the tools you use are less important than the underlying operating system you build. A flexible, layered infrastructure is the key to a future-proof strategy that doesn't require starting over with every new tool.
A recent collaboration with a technological unit in the IDF highlights how shared effort and a focus on the journey are essential to building a leading AI Nation.
Between Israel’s Memorial Day and Independence Day, I found myself reflecting on a core idea: a nation's strength comes from its builders choosing to serve. This isn't just a theory; in Israel, it's a living ethic that underpins our resilience.
Science fiction often portrays AI threats as physical robots, but the more immediate danger lies in distributed agentic networks. These systems are evolving faster than our understanding, creating a critical security challenge for governments worldwide.
Andrej Karpathy recently captured the dual nature of today's AI agents, simultaneously brilliant experts and surprisingly fragile children. This paradox perfectly describes a key challenge we face in the field.
A fascinating open-source project, TradingAgents, reveals a powerful architectural pattern for AI-driven decision-making that can be applied far beyond finance, including in the public sector.
The White House's new AI framework offers a clear policy direction, prioritizing national alignment and a light-touch approach. Here are my key takeaways on its strategic implications for innovation and governance.
NVIDIA's new NemoClaw platform signals a major shift for AI agents, moving them from experimentation to enterprise-ready workflows with a focus on security and governance.
A recent White House report poses a critical question for national leaders: On which side of the AI-driven "Great Divergence" will your country find itself? The answer may depend more on adoption than invention.
While the AI conversation focuses on models, infrastructure pioneer Akamai is building a distributed inference platform that could reshape how AI operates in the physical world.
Companies that rushed into AI-driven layoffs are now making a U-turn, discovering the technology isn't mature enough to replace human nuance and problem-solving. The data shows a clear trend of executive regret and quiet rehiring.
A policy proposal aims to make tech giants fund their own AI energy infrastructure, but the complex financial instruments being used could inadvertently shift the risk back to the public.
The open source community is a powerful evolutionary force. The recent explosion of projects around OpenClaw demonstrates how organizations can leverage collective innovation.
The fear that AI will eliminate jobs overlooks a key economic principle. Jevons' Paradox suggests that increased efficiency often doesn't shrink industries, but scales them by unlocking new demand.
I'm thrilled to announce the launch of Binom, an AI-based system designed to streamline the work of Israeli government committees, marking a significant step in leveraging AI as a force multiplier in the public sector.
While I lead public sector AI initiatives by day, the most meaningful impact of AI I've seen is at my own kitchen table, where my children and I build personalized learning apps together.
I recently witnessed a major milestone for public sector AI in Israel, as the 'Binum' project officially began integrating AI tools to enhance the core work of government committees.
Israel has launched its new national AI supercomputer, but the real story is the strategic approach to its use. By treating compute as a shared national asset, we are building a foundation for the entire ecosystem.
A recent collaboration between Israel's Hailo and Raspberry Pi highlights a crucial shift: Israel is becoming an AI Nation not just through apps, but by building foundational AI infrastructure.
Recent rumors of NVIDIA's potential acquisition of AI21 Labs for $2-3 billion signal a pivotal moment for Israel's AI ecosystem, raising key questions about talent, infrastructure, and long-term independence.
The paradigm of software development has shifted from writing code to orchestrating AI agents. To stay competitive, we must rethink our tools, workflows, and how we measure success.
A common fear is that AI will replace jobs. Drawing on a comment from NVIDIA's CEO and my own experience in healthcare AI, I argue the reality is far more positive: AI automates tasks, freeing experts to focus on high-impact work.
At a recent AI conference, we replaced the standard corporate booth with a sofa. The conversations that followed revealed a powerful vision for collaborative innovation in public sector AI.
A leading AI scientist says we have until 2030 to decide on AI autonomy, but I believe we must act now. To manage AI agents securely, we must stop treating them like tools and start managing them like employees.
A recent meeting between Donald Trump and NVIDIA's CEO offered a critical lesson: fragmented AI projects cancel each other out. A unified strategy is essential for creating compounding value and achieving real scale.
The United States' new Genesis Mission validates the unified platform approach we're pioneering with Israel's GOVAI initiative, highlighting a global shift towards collaborative AI development.
A looming Tesla shareholder vote on Elon Musk's pay package isn't just about compensation; it's tied to audacious goals, including deploying one million humanoid robots, that could redefine our AI-driven future.
A scene from the 2005 film Coach Carter offers a powerful lesson on AI leadership. Building a true AI capability means seeing beyond a narrow job description to connect all the pieces of the organizational puzzle.
The coming world of AI agents promises massive productivity gains, but only if we first solve the complex, 'boring' challenges of identity, authorization, and governance.
A chance encounter with a determined entrepreneur during a morning dog walk served as a powerful, real-world reminder of what true motivation looks like.
I recently attended a Google event in Tel Aviv packed with impressive live demos, highlighting four powerful AI innovations for developers and creators that you might not know about.
At a recent AWS Hackathon, we presented a new component for the GOVAI platform designed to turn static documents into living, queryable knowledge capsules for government employees.
Adopting AI in large organizations creates a natural tension between innovation and security. The solution isn't just another tool, but a fundamental commitment to communication between teams.
The era of simple prompt engineering is evolving. To build complex AI, we must shift to Context Engineering, the discipline of architecting the complete world of knowledge an AI needs to deliver robust, scalable solutions.
Recent rumors about mandatory AI adoption at tech giants miss the point. True AI integration comes from empowerment and trust, not from threats tied to performance reviews.
Google's launch of the open-source Gemini CLI isn't just about sharing code; it's a brilliant strategy to turn the developer community into co-creators and the world's most effective beta testers.
I recently spoke to the Data Driven Government community about my past work on a national health AI platform, and how today's GenAI tools could revolutionize every stage of such a project.
I recently spoke with product managers at the Israel National Digital Agency about how Generative AI is fundamentally changing product development for public services.
AWS has introduced the Agent2Agent (A2A) protocol, enabling Amazon Bedrock Agents to communicate and collaborate seamlessly. This is a significant step for building complex, multi-agent AI systems.
AI's impact on jobs isn't just about cuts; it's a two-phase reset. I see an initial efficiency wave followed by a reinvention wave that presents a major opportunity for leaders who plan ahead.
The recent AWS Summit in Tel Aviv made it clear that the era of AI experimentation is over. The tools and infrastructure are now mature enough for widespread, intentional adoption.
The rapid competition between major LLM providers presents new opportunities for efficiency and cost savings. It's crucial for leaders to regularly reevaluate their AI stack to capitalize on these shifts.
OpenAI's reported $6.5 billion acquisition of Jony Ive's design firm is a powerful reminder that technical accuracy alone doesn't win. True value in AI comes from a focus on human-centered design, user experience, and a strong ecosystem.
While major AI releases often grab the spotlight, Mistral's recent open-source model, Devstral Small, offers a powerful, accessible tool for developers and researchers to drive real-world innovation.
The Israel National Digital Agency is hiring a senior product manager to build advanced data and AI infrastructure, a role with a genuine impact on public services.
A simple analogy from the PydanticAI documentation reminds us of a crucial principle in AI development: always choose the right tool for the job.
As someone who follows AI news closely, I've noticed Google's recent, accelerated pace of releases. A quick look at their updates confirms they are shipping new features and models at a truly remarkable rate.
A recent security post was a powerful reminder of the need for cautious innovation, especially as we push for AI adoption in critical sectors like public service.
I often see organizations struggle with ineffective documentation. The solution is to recognize our new dual audience: to build for the future, we must write for both people and AI.
Effective cloud financial management is more than just saving money, it's about building a culture of transparency and accountability, which is especially critical for unpredictable AI projects.
Google's new Agent-to-Agent (A2A) protocol is a pivotal step towards a future where different AI agents can collaborate, creating an ecosystem of cooperation rather than competition.
Shopify's recent memo on AI sets a powerful standard, but true success requires more than ambition, it demands a supportive strategy that empowers the entire team.
Major open-source releases like Meta's Llama do more than just introduce new capabilities; they energize the entire AI ecosystem, creating a powerful ripple effect that accelerates innovation across the industry.
A recent visit to the Volcani Institute revealed impressive AI applications in agriculture and affirmed a shared vision for centralized government AI infrastructure.
In today's climate, the most promising Insurtechs are focused on cost optimization. I recently saw a powerful example from Israeli company umAI, which uses AI to strategically streamline motor claims processing.
Many companies rush to hire data scientists, but this single role is not a silver bullet. True data-driven value requires building a supportive ecosystem.
A look at the critical role of the Data Developer in making AI initiatives successful, and why they are essential for freeing up your data scientists to do their best work.
Areas of current research and thinking that will inform future writing.
How organizations move beyond pilots to repeatable AI capabilities that integrate with existing workflows.
Why agentic AI requires permissions, orchestration, evaluation frameworks, and operating models, not just smart models.
What complex public institutions teach us about adoption in regulated and high-stakes environments.
Insights from national-scale healthcare AI, including sensitive data handling and real-world deployment.
Connect with me on LinkedIn, or reach out for speaking, advisory, or strategic collaboration.