Well played

I like LLMs but sometimes they take you off on the deep end and you end up spending hours analysing something for very little outcome - and often with large opportunity cost. Sometimes just better to do the analysis yourself.

This seems like the best answer. There is now a much more powerful incentive to get it right.

This history map shows the US bombing in SE Asia during the Vietnam War from 1965 to 1975. Blue dots are 1000 lb bombs or larger. Source: buff.ly/1KXSRrd

The location of the Strait of Hormuz on google with the opening hours shown at Mon-Fri 9.30am - 4pm

Someone has actually done this on Google 😂😂👏

Stuff like what's features on 16colo.rs/pack/corrupt...

I always grin when I use @anthropic.com Claude Code because it uses ANSI/ASCII art - something I myself used to make back in the early 90s as a kid when I ran a Bulletin Board System!! #dataSky, #dataBS #Claude en.wikipedia.org/wiki/ANSI_art

Well....perhaps that ship has already sailed :-/

Maybe the best time was when ChatGPT hit the market, and the second best time is now?

Need to ruminate a LOT more on a lot of this. Is a new, or at least different, role being created here, advocating for responsible use of these types of tools? Tight visual guidelines? A strong return to fundamental visual analytics skills?

Still think it's somewhat inevitable we'll end up using tools like Claude Design+Code to prototype faster, but perhaps there's a way to not lose the other thing.

And I have to admit: I've already seen some pretty awful prototyping in other platforms doing this stuff (Figma Make, looking at you). It does make me inclined to credit that we're just increasing the potential size of a future Dashboard Graveyard.

I do agree that the process of sitting with a client, teasing out what they want to achieve through thoughtful iteration, *is* often quite valuable.

Getting past my initial excitement about speeding up a process (prototyping can be laborious, particularly if you happen to do it in the BI tool itself, which I've seen done), I do find myself nodding along with a lot of what you're saying.

Appreciate the sentiment @datarocksnz.bsky.social! and you're raising some pretty interesting points that are making me think a bit more critically about the place of such tools in the wider BI/analytics toolchain.

I should say, I'm enjoying the discussion and glad to be back and engaged on BlueSky!

Haha, rant away!! :)

Thats an interesting piece for sure (which I've saved for a deeper read). I will just say that my original article is just evaluating the practical uses of a tool - I'm not advocating for the mass firing of all analytics professionals and the scrapping of visual design fundamentals here...

I would say though, I've seem a huge amount of very bad dashboards in production pre-AI that obviously haven't been through the problem triage process. :/ It's not a new thing- but we may have to deal with this at scale.

I like this point and I agree with it - perhaps it will make the 'craft' of the dashboard more important or valuable? We may have to wade through a huge amount of slop to get there though :/

The Four Big Bets Reshaping BI When data display commoditizes, things get weird.

I thought this post by Ryan Dolley on the 'Four Big Bets Reshaping BI' is very well written and pretty much captures how AI is creating a decision point for all major business intelligence platforms. Very worth reading. #dataBS #DataSky #datafam superdatablog.substack.com/p/the-four-b...

I've been a bit lax on Bluesky recently, but it seems the data community is still here! That's good!

You'll note in my quick review, Claude didn't get it anywhere right the first time - despite a fairly 'good' prompt, and I had to basically go through a back and forth with it. Like I would with any junior visual analyst. That's how I think of it.

I sympathise with this view somewhat - there is a huge danger of a production line of terrible dashboards. But, to use a metaphor, one can use a paintbrush to paint an image that resembles something that a toddler might paint or I might paint the Mona Lisa. In the end, AI is simply a tool.

Hey guys, original author here - great conversation. I guess @insightsmachine.bsky.social what I meant by that is that it resembles the work of a competent developer - but I do acknowledge, this stuff does require peer review/quality control. My article was simply a quick 'hot-take' :)

Are My Dashboarding Days Numbered? An Early Test of Claude Design Claude Design just launched. Here's what the design-to-code pipeline actually does in the world of visual analytics, business intelligence and dashboarding.

Are my dashboarding days numbered? I've put Claude Design through its paces on visual analytics (@anthropic) - and here's my write up of the experience . thedatavist.substack.com/p/are-my-das... #dataviz #datafam #databs #datasky

Hey @motherduck.com - can you record and publish your excellent webinars! I can't find the latest one on Dives and I really want to watch it. Unfortunately, being in Australia, it's difficult to get up at 2AM to watch them live.

I asked a Claude-powered four AI agent team to debate the future of higher education using the Socratic method. Here's how it works and why it produces sharper strategic insight than a single prompt. thedatavist.substack.com/p/kill-the-l... cc: @anthropic.com #DataSky #DataBS #AI #AIAgents

Will Ashcroft is good. But is he great? Two premierships and two Norm Smith medals at 21 is already an outstanding AFL career. But how does Will Ashcroft compare to the competition's elite midfielders? Will Ashcroft is considered one of th...

Did a deep dive on just how good Will Ashcroft is and what he might have to do to be a Top 5 midfielder in the #AFL.

roar-metrics.ghost.io/will-ashcrof...

Literally every text message my dad has sent me on a smart phone has been with the SLAM effect.

That’s like a decade of SLAMz

Thanks for the shout out Robert!