Curated Distilled articles at the intersection of Data and the Web
As Agentic AI applications transition from experimental prototypes to production scale, finance and engineering teams are colliding over a new reality: the cost of autonomous operations. This has given rise to the concept of Tokenomics—the microeconomics of autonomous AI execution. Unlike traditional Generative AI applications where a user inputs a single prompt and receives a single response,…
In an earlier article in 2024 I wrote about red teaming AI systems through an adversarial lens- how inputs can be crafted to expose weaknesses, and how thinking like an attacker reveals hidden failure modes. That framing still holds, though it needs to be extended now with advent of more powerful reasoning models and agentic systems. These systems don’t result necessarily in binary outcomes - safe…
Next industrial revolution Recently as part of the Chinese lunar new year celebration there was a showcase of there was a showcase of humanoid robots performing complex, coordinated Kung Fu movements—something that was not feasible even a year ago due to limitations in control, simulation fidelity, and edge inference. This clearly marks the brewing of a new industrial revolution. In this article…
In the last post we discussed context memory of agents. Now, this memory is as good as what goes in. Garbage in is garbage out. And context poisoning is one important type of risk that we will delve into here. What is Context Poisoning? Context poisoning is a high-severity security threat caused by manipulation of upstream data. Note that this does not require access to the model itself, so this…
Most agent failures aren’t model failures - they’re memory failures. Hallucinations, task drift, irrelevant responses, and runaway token costs often stem from one root cause: poorly designed memory architecture. Memory Architecture is a foundational concept of Agentic AI, allowing the systems to retain context, learn from past experiences, and make informed decisions over time. Take the example of…
Prompt Engineering as a practice came up with the statement- “English is the new programming language”. It indeed did bestow some powers for doing many tasks with Language models. However in the last one year or so things have started to change in some interesting ways, with a new discipline called “context engineering”. Let me break this down a bit. AI agents and models need context to perform…
In part 1 of LangSmith observability post we had introduced how to do basic setup and view basic trace information in LangSmith portal. Now lets extend this concept to a very important aspect of Responsible AI called Audit trails. Audit trail is about keeping records of the AI application’s run, so that any 3rd party governance auditor can verify the details of any run, safety threshold…
AI governance that we see today across the world is likely influenced by the history and collective experience of the people in the regions.
Working with Large Language Models (LLMs) can often feel like a black box. You send a prompt and get a response, but what happens during that interaction? Just as software engineers rely on trace logging to debug complex applications, developers building with LLMs need a clear way to see what’s happening at runtime. This is where a powerful tool like LangSmith becomes essential.
What is MCP? Model Context Protocol (MCP) is an open standard from Anthropic for connecting AI assistants to data including content repositories, other systems where data resides, business tools, and development environments. Imagine you have “N” applications (chatbots, coding assistants. or data analysis tools). And you may have “M” different external tools or data sources (DBs, File systems,…