Akul Kailwoo

One of the biggest challenges I see with scaling LLM agents isn’t the model itself. It’s context. Agents break down not because they “can’t think” but because they lose track of what’s happened, what’s been decided, and why. Here’s the pattern I notice: 👉 For short tasks, things work fine. The agent remembers the conversation so far, does its subtasks, and pulls everything together reliably. 👉 But the moment the task gets longer, the context window fills up, and the agent starts forgetting key decisions. That’s when results become inconsistent, and trust breaks down. That’s where Context Engineering comes in. 🔑 Principle 1: Share Full Context, Not Just Results Reliability starts with transparency. If an agent only shares the final outputs of subtasks, the decision-making trail is lost. That makes it impossible to debug or reproduce. You need the full trace, not just the answer. 🔑 Principle 2: Every Action Is an Implicit Decision Every step in a workflow isn’t just “doing the work”, it’s making a decision. And if those decisions conflict because context was lost along the way, you end up with unreliable results. ✨ The Solution to this is "Engineer Smarter Context" It’s not about dumping more history into the next step. It’s about carrying forward the right pieces of context: → Summarize the messy details into something digestible. → Keep the key decisions and turning points visible. → Drop the noise that doesn’t matter. When you do this well, agents can finally handle longer, more complex workflows without falling apart. Reliability doesn’t come from bigger context windows. It comes from smarter context windows. 〰️〰️〰️ Follow me (Aishwarya Srinivasan) for more AI insight and subscribe to my Substack to find more in-depth blogs and weekly updates in AI: https://lnkd.in/dpBNr6Jg

٢٢١

٥١ تعليق

When building an AI chatbot to retrieve order history in an e-commerce setting, consistency matters. This is why we set the model temperature low — to reduce randomness and deliver reliable, repeatable answers. LLMs are inherently probabilistic. Temperature controls how likely the model is to choose less probable outputs. Lower temperature implies more determinism; higher temperature allows for more creativity. It’s a key parameter when designing AI agents for enterprise applications. For example, consider this order status query: User: “What’s the status of my last order?” Response 1: “Your last order for a wireless mouse was delivered on July 28.” Response 2: “Looks like your recent order arrived a few days ago. Let us know if you need anything else!” Both are correct — but in enterprise workflows, variation like this can cause confusion. That’s why tuning model temperature is critical. A few seats are still available in Building AI Agents in Enterprise course that Ajay Shenoy and Abhijith Neerkaje are teaching. Link to register is in the comments below.

٦٩

١ تعليق واحد

Perhaps the second most interesting data platform trend this year is around pricing models. 1. Fabric, once boasting a predictable bill, has now introduced some on-demand elements in addition to still needing a Fabric subscription. For clarity: I am not critizing the value this could bring to Fabric customers, but the loss of the “one predictable bill” selling point I think is significant, specially if that continues to become more of the norm. 2. Snowflake’s Gen 1, Gen 2, and data volume based pricing is also creating a lot more “variety” which in itself is not bad, but decreases the clarity of the otherwise clear pricing that I’ve publicly praised them for. 3. If you’ve been a Databricks customer for a while, chances are pricing still remains one of most complex things about Databricks. This is mainly fueled by legacy compute and the concept of “Serverless”. But in my interactions with newer cusfomers, generally compute pricing is clear and becomes less of a stumbling element. Fun times!

٤٣

١٧ تعليق