Years ago I attended a great session by Brian Mackey discussing Revit’s journal files, the cryptic text logs that writes down everything you do in the software. I think it was RTC (RIP).
And while I learned a lot, cracking into the journals was not a frequent activity, so that knowledge was pushed out of my brain over time. Probably to make place for some cryptic bit of useless information. (Futurama quotes are NOT useless)…
While the power of the journal is undeniable, its inscrutability is as well.
But as I have been doing with a lot of random items these days, I tossed one into a new chat with Claude (C, as I call it around here).
The challenge: We had a user that swore they did XY and Z, but things just didn’t add up. It would have been almost impossible for the events to have unfolded the way they were described, and if they had, there was a major issue with Revit.
I knew the journal would have tracked what happened: what buttons were clicked, what activity was cancelled, etc. But I was not awesome at reading the journal files.
Good news! The computer does a good job at reading computer jargon!
It turns out there was just enough information already published on the web for C to have a basic understanding of the journal contents and with some nudging it gave me a timeline of user actions that were a very different story than what we were told.
Now, my goal here was not to catch someone in a lie… to point out how they were wrong… I cannot tell you how many times I will swear that I completed a task, when in fact, I never started it. This is just how humans work. We are busy and our brains can conflate old and new activities.
No, my first goal was to make sure I didn’t need to open a ticket with Autodesk… check.
My second goal then was to calm down the PIC. Again, not by pointing fingers or publicly shaming someone, just by saying “hey, I know they said they did this, but I bet they got confused and actually did that. Oh, and here’s all the evidence of a running activity log that backs up my theory.”… check.
Third and final goal was to look for a training opportunity. Did we not have our production staff prepped and fully understanding the tool they were using… still in progress.
C did have some hurdles, so if you go down this road I suggest you asking “are you sure?” in a manner that works for you; maybe some prior instructions, maybe an in-chat nudge. Some initial information was wrong, and my skepticism led C to double check and get me the right information.
Make sure your AI of choice is reading the entire provided journal file. There continue to be some odd cache rules that I can’t get my head around. But the first read through did not get the entire text.
Finally, do some manual spot checking. I didn’t know the journal jargon for “sync with central,” but I asked C to show me what line I could find that activity on, and why it thought that jargon meant that specific command. A quick web page review and peek in Notepad++ was a good double check.
This is certainly a tactic I will use again, ideally in the same C Project, that way each chat shares the same context and builds on what the others have already worked through.
And like I said, it’s not to point to someone and say “you’re dumb and wrong,” because we all make mitsakes.
Design Tech Field Guide 