Kadu Anastacio

Data & AI Leader | Microsoft Fabric • Databricks • Purview • AI | Daily takes on the modern data stack | 30 years making data useful, from Sydney

[Kadu Anastacio @kadurp]

Full carousel on LinkedIn 👇
https://t.co/Aq8Sy90zmv

[Kadu Anastacio @kadurp]

Most data loss happens at the edges, not the core. Purview's strength is layering endpoint, browser and network controls so there's no gap for sensitive data to slip through. Swipe through for the three layers that work together 👇

[Kadu Anastacio @kadurp]

Source article 👇
https://t.co/VvpQQlMm2c

[Kadu Anastacio @kadurp]

The Power BI Dataflows Gen1 to Gen2 upgrade wizard is finally here. Automates connection mapping, query migration, and handles incremental refresh. Still in preview, but removes the biggest friction from your migration path.

Continued in the comments 👇

[Kadu Anastacio @kadurp]

Full post on LinkedIn 👇
https://t.co/whKn4BzaJi

[Kadu Anastacio @kadurp]

Full post on LinkedIn 👇
https://t.co/jylPzZePac

[Kadu Anastacio @kadurp]

OneLake URLs just gave Power BI developers native file referencing. No custom connectors, no base64 encoding—just direct paths to images, JSON, and config files stored in Fabric. Chris Webb's walkthrough shows how simple it actually is.

Continued in the comments 👇

[Kadu Anastacio @kadurp]

@nanoguardbrasil @lordivan22 Olha mais adiante q vc vera uma cenoura pendurada na vara.

[Kadu Anastacio @kadurp]

Source article 👇
https://t.co/pUQdaq0Cvp

[Kadu Anastacio @kadurp]

Small language models are faster, cheaper, and often better at focused tasks than the big ones. The trick isn't replacing large models—it's starting with the smallest one that works. Most teams are overpaying without realising it.

Continued in the comments 👇

[Kadu Anastacio @kadurp]

Full post on LinkedIn 👇
https://t.co/zeAJs5VJWU

[Kadu Anastacio @kadurp]

Source article 👇
https://t.co/qOE2V7fCBs

[Kadu Anastacio @kadurp]

Most organisations use AI as a research assistant. But the real shift is already here: AI that doesn't just help you decide—it decides and executes. Three modes are running in parallel, and your infrastructure needs to support all of them.

Continued in the comments 👇

[Kadu Anastacio @kadurp]

Full post on LinkedIn 👇
https://t.co/sx8DnMCWye

[Kadu Anastacio @kadurp]

Full post on LinkedIn 👇
https://t.co/XRoHFilOXH

[Kadu Anastacio @kadurp]

Your agent isn't failing because the model's not smart enough. It's failing because it doesn't know what it needs to know. The real question: are you building retrieval pipelines or governed memory systems with clear accountability?

Continued in the comments 👇

[Kadu Anastacio @kadurp]

Source article 👇
https://t.co/JZaL5LOWlv

[Kadu Anastacio @kadurp]

RAG hallucinations don't come from bad prompts. They come from messy context. Query transformation, smart retrieval, reranking, and curation—boring plumbing that stops failures upstream before the LLM ever sees them.

Continued in the comments 👇

[Kadu Anastacio @kadurp]

Full post on LinkedIn 👇
https://t.co/GNVVDj8iXS

[Kadu Anastacio @kadurp]

Full post on LinkedIn 👇
https://t.co/olDiW5nCRL

[Kadu Anastacio @kadurp]

Data platforms are being rebranded as agentic control planes. Sounds clever until you realise most orgs haven't even unified customer identity across systems yet. The governance layer needs to be real, not aspirational, before agents can do actual w…

Continued in the comments…