Technology Explorations in Data

LLMs & Chatbots become messy fast once you move beyond simple prompts. Meet the latest concept to help you move towards Agentic AI: Skills!As conversations grow longer and agents start reasoning in loops, context windows fill up with noise, increasing the risk of confusion, brittle behavior, and hallucination. All of this makes it harder to work with agents.AI agent skills in Claude Code offer a lightweight way to solve this. Instead of rewriting prompts or stuffing everything into a single conversation, you can package repeatable workflows into modular, reusable components that load only when needed.In this episode, Jesus walks us through how skills work in practice. We explore the shift from prompt engineering to context engineering, show how progressive disclosure keeps context clean, and demonstrate two real examples: an Explain Code skill that structures educational breakdowns, and a PR Review skill that forks context, limits tool permissions, and uses CLI commands to analyze pull requests. We also discuss when to use skills versus MCP and how to design concise, non-brittle skill definitions.

Surprised by your Azure Log Analytics bill? We found that you can cut costs by 60%!Azure Log Analytics costs often take up 20% or more of a cloud bill, even though most teams only check logs when something breaks.Azure’s default analytics logs are powerful, but they’re also expensive and often unnecessary for day-to-day log inspection. Switching application logs to Basic Logs can reduce Log Analytics costs by up to 60%.In this episode, Niels walks us through a real customer case where logging costs dropped by thousands per year. They explain the difference between Analytics, Basic, and Auxiliary logs, show when Basic Logs are sufficient, and discuss practical setups using Azure Container Insights and FluentBit. This includes building a custom FluentBit plugin in Go as well as real-world gotchas like missing short-lived pods and why dynamic credentials matter.

Code reviews can be a pain: rushed approvals, “rubber stamping,” and bugs slipping into production. With AI-assisted coding accelerating how much code we produce, the review bottleneck becomes even more real.In this episode of Technology Explorations at Dataminded, Hannes De Smet (data & platform engineer) shows Jonny Daenen what he learned after exploring AI code reviewers and demos CodeRabbit and Sourcery on a real code change.You’ll see:How AI reviewers work as a pre-flight check (before opening a PR) and in-PR (via GitHub integration)What kinds of issues they catch well (e.g., type mismatches, logic errors like list handling, division-by-zero)Where they struggle (e.g., noisy “PII exposure” warnings without enough context)UX differences in IDE integrations (Cursor/VS Code), “apply fix” vs “fix with AI,” and why context still mattersLinks:CodeRabbit: https://www.coderabbit.ai/Sourcery: https://www.sourcery.ai/

Emil Krause and Jonny Daenen explore the claim that speech-to-text makes you three times as fast. Emil shows the functionalities and benefits of Super Whisper and Whisper Assistant, two innovative speech-to-text tools designed to enhance productivity. They discuss the installation process, user experience, and the accuracy of these tools. Emil shares insights on how these tools can streamline workflows, particularly for those who frequently interact with AI, and emphasizes the importance of context in dictation. It turns out that SuperWhisper gives a much better accuracy when it comes to technical terms.Whisper Assistant: https://marketplace.visualstudio.com/items?itemName=MartinOpenSky.whisper-assistantSuperWhisper: https://superwhisper.com/Note: This video is not sponsored or affiliated with Whisper Assistant or SuperWhisper.

When working across multiple codebases, your AI assistant only sees one at a time; you lose context the moment you switch repos. Cursor Workspaces aims to fix this by letting you pin multiple repositories into a single indexed workspace.In this episode, Jonny and Emil demo the feature with a practical scenario: an app running out of memory that needs both a Kubernetes deployment change and a Terraform update in a different repo. Along the way, they test whether the AI can figure out cross-repo dependencies on its own, and discover why human involvement still matters.You’ll learn how to:Set up Cursor workspaces with multiple projects.Let your agent access and manipulate multiple codebases.Leverage best practices from previous projects.

In this episode, Jonny and Tarik show how to build a fully functioning RAG-based AI agent using MindsDB, turning internal data from Postgres, Slack, or Google Drive into a queryable knowledge base powered by semantic search.You’ll see how to:Convert internal documents into a semantic knowledge baseMake Google Drive & Slack data queryable in minutesBuild a custom AI Agent on top of your knowledge basesRun MindsDB locally with DockerUse SQL to configure agents, connectors, and knowledge basesExpose your agent through a simple API for app integrationWe also touch on:Chunking & embedding strategiesLocal vs. cloud LLMsHow MindsDB compares to full ETL approaches

Move your Google Drive documents straight into Postgres using Python and PyAirbyte. In this Technical Explorations episode, Jonny and Tarik from Dataminded show how they ingest internal meeting transcripts (Facts & Breakfast, Learning Over Lunch) from Google Drive into a relational table, ready for querying and AI use cases.You’ll see how to:Configure PyAirbyte to read from a Google Drive folderAuthenticate with a Google service account (JSON key)Convert Airbyte output into a clean pandas DataFrameLoad the processed data into a Postgres tableDiscuss performance limits, API rate limits, and batchingReflect on when PyAirbyte is great for PoCs vs. production setupsWe also touch on:How many connectors Airbyte offers and what PyAirbyte can reuseTrade-offs of code-first ingestion vs. point-and-click UIIdeas for the next step: using MindsDB and LLMs to query this knowledge base

Emil Krause & Jonny Daenen explore how to accelerate dbt development by integrating MCP (Model Context Protocol) with Postgres and Cursor. Emil demonstrates how to solve a database bug by allowing AI agents to interact directly with databases and his code. Even though this approach is very effective, it remains important to understand the data, even when using advanced tools.https://github.com/datamindedbe/demo-technology-exploration/

In this special episode on the AWS outage, Stijn De Haes explains to us what happened during the AWS October 2025 Outage. He then zooms in on the limited effect it had on Dataminded and its product Conveyor. And finally he gives 4 tips on how to protect yourself from this kind of outage.👉 Link to the blog post: https://hubs.li/Q03PQKrR0

In this conversation, Jonny Daenen and Pierre Crochelet explore the Model Context Protocol (MCP), a framework that enhances AI assistants by allowing them to perform various tasks through tools, resources, and prompts. They discuss the architecture of MCP, how to build an MCP server, and the developer flow for creating tools. The conversation also touches on the compatibility of MCP with different AI agents and the user experience, highlighting both the potential and limitations of the protocol.👉 Link to the demo code: https://github.com/datamindedbe/demo-technology-exploration