Google doesn't understand Panorama and neither do I

rtln · June 25, 2026, 1:51am

I have a Text editor field in a form.

After data is entered into that field, a procedure executes

Starting this afternoon, that procedure (after years of working) started throwing an error.

I want to edit that procedure or at least watch it fail, and I CAN’T SEE THE PROCEDURE.

I asked Google, how do I see the procedure that executes after data is entered into a field in a form in Panorama, and it told me,

“Steps to View the Text Editor Code
Switch to Graphics Mode: Press Command-Option-G or choose Form > Graphics Mode.
Select the Text Editor: Click once on the specific Text Editor object to select it.
Open Object Properties: Open the Blueprint or Properties panel on the right side of the screen.
Click the Procedure Tab: Look for the icon resembling a magic wand at the top of the properties panel.
Inspect the Code: The text area below will display the exact anonymous procedure that runs when data entry concludes.”

Everything performs except the last step. Clicking the Wand does not show me a procedure, It shows me a whole pile of options pertaining to text editing, and none of those options takes me to the procedure.

I’m feeling quite stupid because I wrote the procedure that executes after the field is edited, and I do not remember how I got to it.

Any advice?

admin · June 25, 2026, 4:07am

I’m not a fan of LLM’s (so called “AI” models), this is an example of why. Google’s AI answer was oh so confident and pretty close, but no cigar. The panel you want is not the magic wand, but the “scroll” icon to the left of the wand.

More details about editing object code can be found on this help page.

RAmeeti · June 25, 2026, 5:07am

I have been using Claude for about 40 hours in the last week and am very impressed at the quantity of work that can be accomplished. It is always right? No. But it is fast. It takes a knowledgeable person to create a good end product with lots of testing, but the efficiency is off the chart. I have accomplished what would have taken multiple months or more in those same hours. I asked Claude about Panorama and here was the conversation…

An off topic question. I use a database called Panorama by ProVUE. It has an extensive programming language. Is there anything that the developer of the program can do so that any user who happens to use Claude will benefit by Claude already having read the online documentation of the program? Or other features that might be accomplished by having a hook into the user’s program?

===

Great question. There are a few distinct things the developer could do, ranging from simple to sophisticated:

Training data submission

Anthropic has a process for submitting documentation to be included in future training runs. If ProVUE submitted the full Panorama X documentation, procedure reference, and example code, future versions of Claude would have native knowledge of it baked in. The catch is that training cutoffs mean it would always be somewhat behind the current version, and there’s no guarantee of inclusion or timeline.

A custom Claude integration (most practical)

The developer could build a Claude-powered assistant directly into Panorama X using the Anthropic API. This would work by:

Bundling the full documentation as context that gets sent with every query
Giving Claude access to the user’s current procedure code, field structure, or error state
Letting the user ask questions like “why isn’t this procedure working” and Claude sees the actual code

This is well within reach today and wouldn’t require any special relationship with Anthropic — just the public API.

An MCP server

This is probably the most powerful option. The developer could publish a Panorama X MCP server that users install alongside the Claude desktop app. It could expose tools like:

Read the current procedure being edited
Read the database schema (field names, types)
Run a test procedure and return the result
Look up documentation for a specific function

Claude would then be able to operate with full live context — not just answer questions about Panorama, but actually inspect and help fix the user’s specific database in real time. That’s essentially what your Claude for Chrome setup does for WTD.

A system prompt wrapper

Simplest of all — ProVUE could host a Claude-powered chat page on their own site with the entire documentation pre-loaded as a system prompt. No API complexity for the end user, just a purpose-built “Ask about Panorama” assistant.

The MCP server route is the most exciting because it mirrors what professional IDEs are doing with AI — the assistant sees your actual work, not just generic documentation. Given that Panorama users tend to be power users building real applications, that kind of deep integration would be genuinely useful. You might consider passing this along to Jim Rea at ProVUE — it’s a tractable project for a developer of his caliber.

RAmeeti · June 25, 2026, 5:40am

The Claude agents are EXTREMELY confident. Almost obnoxiously so. And I take no responsibility for their brown nosing.

The agents are what would be described as 18 year old geniuses who have a fair amount of experience but an unlimited knowledge and unlimited enthusiasm. They know Exactly what the problem is after every failure, and they know Exactly what the fix is. Until it fails for the 4th or 5th time. But they do not give up, and they do not lose enthusiasm. I am an extremely particular person about the way I want things done. I do not stop until it is perfect in my eyes. If a normal employee would be working for me, they would quit. Or they would say “I’m really tired of working on this, can we stop now because you are being obnoxiously anal about this.” But not Claude. He will do it again and again until I think it is perfect. And he is 500% better than me at design, and 1000% times faster and forever polite. When I point out poor decisions on his part, he apologizes and takes my advice. He learns the way I like to work and he follows my methods. Yes sometimes I have to remind him of some things, but he apologizes again and is good for a good period of time.

I am anxious for Jim to give it a try as I am sure that a few bugs and a few features that we are all waiting for can be quickly accomplished with a very inexpensive genius that will work at any time of the day with full and complete enthusiasm.

Bottom line, after I fed Claude the online documentation, he made very few mistakes. Very few.

rtln · June 25, 2026, 1:06pm

What you said is the first thing I tried, looking at the tab where the procedure should be, the scroll which lies to the left ot the wand. Problem is, it was blank. There are no words at all, under the “Procedure. Apply” header when I look at the properties of the field in graphics mode. Yet when I edited the field, in Data mode, the procedure I have written to be associated with that field executed - and failed, due to a piece of bad data introduced, and I am not kidding, by a small terrier who has a habit of jumping up on my keyboard for attention. And now that I have fixed the data, the procedure executes correctly. But the procedure tab still shows a blank. It shows blank for all the fields on that sheet, yet all the procedures associated with the different data fields are behaving properly.

I hope that Claude is absorbing all these comments, and can offer relevant output for programmers who are puzzled by this piece of behavior.

rtln · June 25, 2026, 1:29pm

wait wait wait

This is my mistake: the procedure that executes after I edit my field “plus” is to be found by looking at that field on the DATA SHEET, where the FIELD PROPERTIES will show you the procedure I wrote, when you click on the “Code” button at the bottom of “Field Properties. But that procedure is apparently not associated with that data field on the form in question. But it does execute, when you edit the data in the form.

I am often confused by the relation between the data sheet and the forms that refer to that data. Simplest case: ”find” and “morph” behave differently if you access them from the data sheet, or from a form. In Panorama 6 and before, you could be sitting on field in a form and say command F and the “find” window you got was the same as you got when you hit command F in the data sheet.

Hmm

designer · June 25, 2026, 4:24pm

Speaking of AI …

As an aside, for those curious, the “gimmick” in yesterday’s universal crossword was three crossword answers relating to “two drink minimum”. For example, one of the answers was “team sport,” and you can pull out tea and port - two drinks. Another was “chair umpire,” and you can pull out chai and rum - again, two drinks. But the third answer was “pentatonic scale”.

I asked ChatGPT about it, and in the style Robert indicated, ChatGPT pulled out “tonic” and confidently set about justifying getting the “e”, “t”, and “a” from Penta for the answer tea and tonic. Justification after justification was wrong. It even asserted, like some politicians, blatant falsehoods, like the e, t, a, taken in that order, spelled tea.

As a human, it only took a microsecond to see tonic in the first word and ale in the second word.

This whole AI thing, among “civilians,” reminds me of the articles we see every winter of people blindly following their GPS down some shortcut forest service road and getting stuck in the snow.

AI’s are great with, “Give me the best sort algorithm for a list of 1000 items that are mostly already sorted when a new element is added.”

But if you blindly follow the suggestions without a scintilla of discrimination, you may find yourself … stuck in the snow.

I much prefer Claude over ChatGPT. I use it for trigonometry when figuring out angles and lengths of potential ham radio antennas for the backyard. Not that I couldn’t do it myself, but with a fixed amount of time, one decides where to spend it.