I just got back from AI Dev x NYC, the AI developer conference where our community gathers for a day of coding, learning, and connecting. The vibe in the room was buzzing! It was at the last AI Dev in San Francisco that I met up with Kirsty Tan and started collaborating with her on what became our AI advisory firm AI Aspire. In-person meetings can spark new opportunities, and I hope the months to come will bring more stories about things that started in AI Dev x NYC! The event was full of conversations about coding with AI, agentic AI, context engineering, governance, and building and scaling AI applications in startups and in large corporations. But the overriding impression I took away was one of near-universal optimism about our field, despite the mix of pessimism and optimism about AI in the broader world. For example, many businesses have not yet gotten AI agents to deliver a significant ROI, and some AI skeptics are quoting an MIT study that said 95% of AI pilots are failing. (This study, by the way, has methodological flaws that make the viral headline misleading; see link in original post.) But at AI Dev were many of the teams responsible for the successful and rapidly growing set of AI applications. Speaking with fellow developers, I realized that because of AI's low penetration in businesses, it is simultaneously true that (a) many businesses do not yet have AI delivering significant ROI, and (b) many skilled AI teams are starting to deliver significant ROI and see the number of successful AI projects climbing rapidly, albeit from a low base. This is why AI developers are bullish about the growth that is to come. Multiple exhibitors told me this was the best conference they had attended in a long time, because they got to speak with real developers. One told me that many other conferences seemed like fluff, whereas participants at AI Dev had much deeper technical understanding and thus were interested in and able to understand the nuances of cutting-edge technology. Whether the discussion was on observability of agentic workflows, the nuances of context engineering for AI coding, or a debate on how long the proliferation of RL gyms for training LLMs will continue, there was deep technical expertise in the room that lets us collectively see further into the future. One special moment for me was when Nick Thompson, moderating a panel with Miriam Vogel and me, asked about governance. I replied that the United States’ recent hostile rhetoric toward immigrants is one of the worst moves it is making, and many in the audience clapped. Nick spoke about this moment in a video (links in original post). [Truncated for length; full text with links: https://lnkd.in/gQhdiY8B ]
Systems Density. The ROI gap isn’t about AI capability; it’s about implementation architecture. The 95% pilot failure rate exists because too many businesses still treat AI as a feature to deploy, not as a catalyst for true systems transformation. Successful teams recognize that AI ROI compounds when you couple™ technical depth with organizational learning loops: not when you bolt intelligence onto legacy workflows. AI Dev’s team exemplifies this approach: They’re building adaptive systems where AI reasoning is embedded into the business’s DNA, not just automating individual tasks. Which implementation patterns are you seeing that truly separate the 5% of successful AI deployments from the rest?
Thanks for organizing AI Dev x NYC, Andrew. I was there in person and the energy really was buzzing — especially around the very real, enterprise-grade conversations on agents, context engineering, governance, and what it actually takes to scale AI beyond pilots into measurable ROI. As someone building AI platforms in a large corporation, it was refreshing to be in a room of builders sharing what’s working (and what’s not yet) at scale. Appreciate everything you and the team do for the developer community. Also great meeting you briefly and getting a selfie together — a highlight of the day!
This kind of energy around real AI development. When builders come together, progress moves faster. Excited to see how these ideas evolve into practical products. And for anyone exploring AI ads, Nyra AI gives a 14-day free plan https://www.getnyra.ai/billing
Andrew Ng - Loved that real developers were collaborating and driving a genuine agenda, rather than just adding fluff. Unfortunately, the rhetoric against immigrants is already present in the UK as well. It divides society and helps keep the powerful in their place. Sadly, it has become a billion-dollar industry, and the most vulnerable people are the ones paying the price.
There’s something really energizing about hearing this perspective from someone who’s been at the center of every major AI wave. What stood out to me here is the duality you mentioned: AI penetration is still low in most businesses, and yet the teams who are investing the right way are already seeing real ROI. That gap is exactly where the next decade of opportunity lives. 📈 In my own career transition, AI has become the unfair advantage. I’m not a developer, but I’ve been able to build workflows, automations, and insights that would’ve been impossible for me even six months ago. It’s made me realize how much potential exists for people who are willing to take the leap early, get curious, and experiment. Hearing that AI Dev x NYC was full of real practitioners (not just hype) confirms what so many of us are seeing on the ground: this moment really is different. The people building now are shaping the foundation everyone else will stand on later. We're watching the future being built in real time. Love seeing this level of optimism backed by actual technical depth. Excited to keep building, keep learning, and play my small part in what’s coming next. 🤝🏼 Like the way I think? I’m transitioning careers. Let’s chat.
The developer optimism makes sense. But from Europe, the deeper concern is not AI itself, it’s the “natural” political project emerging from its private U.S. infrastructure. Today, the platforms that filter visibility, meaning, and legitimacy are not public, not democratic, and not accountable. They form a cognitive environment that gradually replaces institutional decision-making with private governance. This is not just economics. It is the quiet substitution of democratic authority with privately owned architectures of thought. In 1975, here in Europe, we would have called this what it is: a form of technological authoritarianism (if not more simply and clear techno-fascism). Today we call it “technology” simply because we’ve become accustomed to it. And there is a principle no society can afford to forget: No corporation and no private actor should ever control the flow of meaning for an entire civilization. If they do, they are not innovators. They are merely satraps. The real question for the future is not how AI will scale, but who has the mandate to shape the cognitive world we all inhabit.
What would be really compelling is a curated list of concrete examples of where Agentic AI and similar tech is being used in production on an ongoing basis and delivering value without succumbing to major performance, security or similar issues. Everybody hears lots of stories of people getting impressive results in limited settings. I suspect what's needed at this point, due to the insane build up around AI both hype and investment wise, examples that are truly game changing for the business. Being merely very useful may not be enough. My personal experience is that it can be game changing but the tech by itself is not enough - the data, the business process within which it operates, the system architecture, and many other aspects are vital to making it successful - real and very hard work is needed to get it there and it's not being sold as such.
I interpret that 95% pilot failure statistic differently. I don't think it’s a methodology flaw, (a claim currently asserted without evidence), I think it’s a reflection of the Unit Economics of Safety. The deep technical understanding you noted is required precisely because we are facing a Guardrail Paradox. To take these probabilistic models from pilot to production, we aren't just deploying code; we are engineering a massive safety stack (20+ layers of filters, RAG verifiers, Agent-to-Agent critics, Knowledge Graphs, etc.). This creates a Safety Tax that often breaks the ROI. We are effectively spending 3x the compute to verify the output as we do to generate it. Furthermore, the Context Engineering you mentioned is the critical bottleneck. We are trying to bridge the Context Chasm, taking models trained on the public 5% of the web and forcing them to execute on the private 90% of enterprise data. The pilots struggle because they are trying to replace Human Middleware, the implicit, tribal knowledge that holds organisations together, with different LLM logic. The developers are bullish because they see the capability increasing. The enterprises are cautious because they see the cost of trust increasing just as fast. Thx