Maxim Anatsko

I've been thinking about impact of agentic development on us humans: agents can scale without limits, while people don't. Need more output? Just spin up another agent. And another, and another. And every one of them sends back summaries, code diffs, questions. All of them land on you. Our working memory can hold only few things at once. More information usually helps us decide better, up to a point. Past it, decisions get worse and every switch between tasks costs you something. Keep that up for months and you will... burnout. Reflecting on the last two years I realized that the new bottleneck is our attention and capacity to process information. Our brains were never designed to deal with this bombardment of inputs and decision-making. To deal with that we need new tools and techniques: ones that let us do more and process less. But it also means we have to learn to trust AI and agents with decisions that were previously in our hands. And for me personally that was the biggest adjustment to make. #AI #AIAgents

Join @marcorus and me live on Wednesday, November 4 at 15:00 UTC to talk about DAXBench v2: how we built it, what the results show, and your questions. We'll stream on YouTube and live here on X too. Set a reminder: youtube.com/live/ct3g7PQiPBQ #DAX #PowerBI #LLM #AI #DAXBench

Opus 5.5 made me a believer again. After a year of Codex, I’m back to Claude. Goddamn, such a good model

DAXBench v2, benchmark for how well LLMs write DAX, is finally here! BIGGER, broader, with more variety and new features like UDFs. Done in collaboration with SQLBI and @marcorus. Technical breakdown on my blog: maxanatsko.com/blog/daxbench… ___ #DAX #PowerBI #LLM #AI #DAXBench

It’s important to give brain idle time from agents.

PBIR CLI 0.9.33 is out. @kurtbuhler and I added the native Gantt chart support and states for batch CF, and fixed the bugs you reported: padding, line chart legends, Input Slicers, bookmarks after a field rename. Thanks to everyone who filed issues. #PowerBI #PBIR

AI makes us faster, but it also changes how much we think we can carry. Saying YES gets easier, and the bottleneck moves from execution to judgment. The next skill is learning what not to take on just because AI made it possible.

Fabric Apps is “the new Power BI”, and yet Power BI got its traction 10 years ago by being a self-service tool first. Fast forward 10 years, and Fabric Apps are React apps requiring a full-stack skill set. Most organizations and Power BI developers simply don’t have that skillset

Who’s coming to FabCon in Barcelona? It was a bit of a last-minute for me, but I’ll be there. If you’re around and would like to meet, hit me up. #FabConEurope #MicrosoftFabric #PowerBI

Power BI Desktop on Mac? After 10 years still no native app. And yet I used Macs for ~10 years while working with #PowerBI almost daily. New video showing how I run Power BI and all of the community favorites without giving up my normal macOS workflow. 🎥 youtu.be/00VKkp9DyfM

I’m taking a break from AI. Next two week no agents, no tokens, all subscriptions downgraded. Only mountains, sea, and books. Vacationing the old school way 😄

Don't know how I slept on this one - building Power BI Custom Visuals. This is the Flow Funnel visual I build in one evening. Nowadays, having good harness, domain and engineering knowledge, and AI at your disposal, changes ROI for building smth like this completely. #PowerBI

I've been using Power BI writeback since day 1. And to be completely honest, I wish I didn't have to. This is where Fabric Data Apps (Rayfin) come in with their flexibility. YouTube video: youtu.be/O2RAWlrLlSk __ #PowerBI #MicrosoftFabric #Writeback #Rayfin

SemanticOps Test Runner v1.1.39 is out, with test stability classification, order-tolerant baselines, steadier run behavior. semanticops.dev/test-runner #SemanticOps #PowerBI #SemanticModels #DAX