569 - Translytical task flows are probably headed where metric sets went. Asked point blank whether they beat a Fabric app at anything, and the one advantage left is pretty thin.
https://t.co/kaGsKRbhqz
569 - These write back vendors charge serious money per user because setting the thing up used to be genuinely hard. With agents and Rayfin in the mix, I'm done paying a third party to do it.
https://t.co/kaGsKRbhqz
569 - Translytical task flows ship write back in the box, and you pay per item you create. A handful of other companies charge for the same capability. Worth asking what that number should actually be.
https://t.co/kaGsKRbhqz
569 - A semantic model does the heavy compute before you ever open the report, with joins and filters already locked in. Views flip that timing completely, and the tradeoff you get back is worth hearing.
https://t.co/kaGsKRbhqz
571 - My agent read the docs page and still couldn't tell which commands run on the CLI versus in the notebook. Good product underneath. The documentation needs work.
https://t.co/zMURrwub2X
#microsoftfabric#documentation#cli#aiagents
571 - Kurt Buhler dropped three Cs on the podcast and they've become my hiring filter. Curious is the one I screen for hardest, because the tooling shifts every week.
https://t.co/zMURrwub2X
#hiring#datacareers#aiskills#powerbi
571 - Picture MCPs disappearing from Fabric tomorrow. The amount of relearning would be ugly. Once you let something in, putting it back in the box isn't really an option.
https://t.co/zMURrwub2X
#mcp#microsoftfabric#aiagents#techdebt
571 - Next article is the Fabric recipe, the agentic meal. Heavy dose of skills, a big cut of MCPs, and the toolkit as the sauces and spices. None of it works solo.
https://t.co/zMURrwub2X
#microsoftfabric#mcp#agentskills#dataengineering
571 - Stop using agents to get the answer about your data. Use them to build the system that gets the answer every time you run it. The machine that makes the machine.
https://t.co/zMURrwub2X
#aiagents#dataengineering#microsoftfabric#datapipeline
571 - My neighbor who can build anything comes over and takes the whole project. I stand there asking what the frame is for. Somebody who's never touched a notebook can now ship artifacts to a lakehouse.
https://t.co/zMURrwub2X
571 - Walked into the hardware store and found an impact wrench sitting next to the drills I'd used my whole life. Great tool. You still have to know how to be the carpenter.
https://t.co/zMURrwub2X
#microsoftfabric#notebooks#aitools#devskills
571 - "Hey agent, cell three is broken, go fix it." The agent reads the error, inspects the cell, patches it in place and runs it again. No more copying code between windows.
https://t.co/zMURrwub2X
#microsoftfabric#notebooks#aiagents#dataengineering
571 - The hack is simpler than it sounds. Tell your desktop agent to install the CLI tool, hand it the docs link, let it fumble through logging in. Then build the skill from what it learned.
https://t.co/zMURrwub2X
#agentskills#cli#claudecode#aitools
571 - Someone called 2023 the start of the conductors era and it stuck. Anyone can tell an agent to build a pipeline. Knowing how to verify it is the part people skip.
https://t.co/zMURrwub2X
#agentskills#aiworkflow#conductorskills#promptengineering
571 - The toolkit is cool and also a little scary. This one belongs with the BI developers who know what they're doing. Random Joe Schmo is not building notebooks this way.
https://t.co/zMURrwub2X
#microsoftfabric#notebooks#governance#biteam
571 - One AI writes the code against our standards. A second one reviews it and rejects the pull request if anything is a copy of code that already exists. Make it a function.
https://t.co/zMURrwub2X
#codereview#aicoding#standards#pullrequest
571 - A client says they don't allow AI use and the conversation ends there. Plenty of companies want it. The real question is whose license you're working under.
https://t.co/zMURrwub2X
#aipolicy#consulting#githubcopilot#datagovernance
571 - Here's the hot take. The Fabric notebook UI is sliding toward optional, and going there might soon mark you as a beginner. Wild thing to say about the default tool.
https://t.co/zMURrwub2X
#microsoftfabric#notebooks#hottake#dataengineering