Welcome to my personal website, a space where I blog and share my musings on academic and professional topics. Dive in to explore my intellectual journey and gain insights from my experiences in various fields!
B2B commerce search has a dirty secret: the ranking algorithm is rarely the problem. When results are bad, teams reach for the fun levers. Boosts, synonyms, embeddings, rerankers, and lately LLM query understanding. I’ve pulled all of those levers myself. Last week I did the unglamorous thing instead: I dumped every product record from four production search indices - 775,051 records across four…
For a while, “prompt engineering” was the name we gave to the craft of getting good results from large language models. It made sense in the early days. Most people were using one-shot interactions, and the main lever really did feel like wording: ask more clearly, add an example, constrain the format, and the model behaved better. That framing is now too small for the real problem. When an AI…
Something interesting is happening in the agentic infrastructure space, and it is not what most people expected. For the past two years, the dominant paradigm for giving agents access to knowledge has been Retrieval-Augmented Generation: embed your documents, store them in a vector database, and let the model query them at inference time. RAG worked. It was good enough. But “good enough” has a…
It started with a single press release. On January 30, 2026, AI startup Anthropic announced 11 specialized plugins for its Claude Cowork agent, empowering it to handle complex workflows in sales, finance, legal, and HR [1]. Wall Street’s reaction was not just negative; it was apocalyptic. In the week that followed, nearly $1 trillion in value was wiped from software and services stocks in a…
What if the best user interface was no interface at all? For decades, we have been trained to navigate a labyrinth of menus, buttons, and settings screens. We learned the language of software. In 2026, that paradigm is finally flipping. Software is learning to speak our language. This is not just about adding a chatbot to a dashboard. A quiet revolution is underway: Software as a Service (SaaS) is…
In the last few weeks, the AI community has been captivated by a project that is not a new model, but a new paradigm. OpenClaw, an open-source personal AI assistant, has exploded in popularity, amassing over 114,000 GitHub stars in just two months [1]. Andrej Karpathy, one of the most respected voices in AI, described it as “genuinely the most incredible sci-fi takeoff-adjacent thing I have seen…
The SaaS landscape is at a critical inflection point. The traditional, human-driven application model is giving way to a new paradigm: the agentic workspace. This is not a distant trend, but a strategic imperative for today. We propose that the next evolution for every successful SaaS company is to become a platform that orchestrates intelligent agents to achieve user outcomes. This transition is…
The concept of Context Graphs , first articulated by Jaya Gupta of Foundation Capital, has rapidly captured the industry’s imagination [1]. The thesis is that the next trillion-dollar enterprise platforms will not be systems of record for data, but systems of record for decisions . For the underlying definition, see my explainer on what are context graphs . The thesis is compelling. But the most…
It’s been over a year since I made Cursor my primary IDE, and it’s hard to overstate the impact it’s had on my work. As a machine learning engineer building conversational AI platforms at Dylog and experimenting with agentic infrastructure on my personal projects, I’ve lived through the evolution of AI-native development. My journey with Cursor mirrors the maturation of the tool itself: from a…
Last week, I wrote about my reaction to Jaya Gupta’s viral post on Context Graphs [1]. The idea of a “system of record for decisions” resonated deeply, framing the evolution of agentic infrastructure from tools to skills to memory. But since then, the conversation has exploded, and it has become clear that the term “context graph” itself is a bit of a Rorschach test. Everyone sees something…
After reading Jaya Gupta’s post about Context Graphs, I have not been able to stop thinking about it [1]. For me, it did something personal: it gave a name to the architectural pattern I have been circling around in the agentic infrastructure discussions on this blog for the past year. I later wrote a more direct explainer on what are context graphs and why the term means more than agent memory.…
If 2024 was the year of AI experimentation, 2025 was the year of industrialization. The speculative boom around generative AI has rapidly matured into the fastest-scaling software category in history, with autonomous agents moving from the lab to the core of enterprise operations. As we close out the year, it’s clear that the agentic AI landscape has been fundamentally reshaped by massive…
The enterprise AI landscape is at a critical juncture. We have powerful general-purpose models and a growing ecosystem of tools. But we are missing a crucial piece of the puzzle: a standardized, portable way to equip agents with the procedural knowledge and organizational context they need to perform real work. On December 18, 2025, Anthropic took a major step towards solving this problem by…
The era of AI experimentation is over. What began as a speculative boom has rapidly industrialized into the fastest-scaling software category in history. According to a new report from Menlo Ventures, enterprise spending on generative AI skyrocketed to $37 billion in 2025, a stunning 3.2x increase from the previous year [3]. This isn’t just hype; it’s a fundamental market shift. AI now commands 6%…
Enterprise AI has a platform problem. The tools to build AI-powered applications exist, but they’re scattered across three disconnected ecosystems—each solving part of the puzzle, none providing a complete solution. This isn’t a “too many choices” problem. It’s an architectural one. Gartner tracks these ecosystems in separate Magic Quadrants because they serve fundamentally different users with…
The AI agent market is experiencing its own big bang—but this rapid expansion is creating fundamental fragmentation. Enterprises deploying agents at scale are caught between two incomplete solutions: Agent Builders and AI Gateways . Agent Builders democratize creation through no-code interfaces. AI Gateways provide enterprise governance over costs, security, and compliance. Both are critical, but…
Just over a week ago, the Model Context Protocol celebrated its first anniversary with the release of the 2025-11-25 specification [1]. The announcement was rightly triumphant—MCP has evolved from an experimental open-source project to a foundational standard backed by GitHub, OpenAI, Microsoft, and Block, with thousands of active servers in production [1]. For readers comparing the protocol to…
Enterprise adoption of AI agents has reached a tipping point. According to McKinsey’s 2025 global survey, 88% of organizations now report regular use of AI agents in at least one business function, with 62% actively experimenting with agentic systems [1]. Yet this rapid adoption has created a critical disconnect: while organizations understand the importance of governance, they struggle with the…
The age of agentic AI is no longer on the horizon; it’s in our datacenters, cloud environments, and business units. A recent PwC report highlights that a staggering 79% of companies are already adopting AI agents in some capacity [1]. As these autonomous systems proliferate, executing tasks and making decisions on behalf of the enterprise, a critical governance gap has emerged. Without a robust…
Anthropic has detailed its disruption of the first publicly reported cyber espionage campaign orchestrated by a sophisticated AI agent [1]. The incident, attributed to a state-sponsored group designated GTG-1002 , is more than just a security bulletin; it is a clear signal that the age of autonomous, agentic AI threats is here. It also serves as a critical case study, validating the urgent need…