FileMaker 21.1.1 fixes the issue where field name changes would break layout calculations. 🙌🏻 I love button bars. But the one annoying thing I don't like about them is the extra work needed to maintain a state with no "active segment" where you don't want it to look "clicked" after the action/script runs. I find that having it set to "None" is not reliable. Setting it to a calculation of `False` works sometimes… but I find that you still need something in your script refresh that object. Even then I've had cases where it just refuses to refresh for some reason. If you have "one simple trick…" that makes that work, make a video for that!

1:15 If you truly want to stir up some controversy and discussion, I have some opinions on the malfunctioning Filemaker Pro universe of naming conventions, which is programming malpractice imho and endemic in FM. There is not one major player who's naming conventions are even close to what users would already intuitively know and understand. I can demo the top 3 templates all with goofball naming conventions. BTW, talking about 'above the screen' not field names. I like where you are going with this series. I've watched a few already.

Is there a link to the prev episode? about context prompts?

I have an odd question. Can a Filemaker Pro file have zero json, if it's Filemaker Pro only. No server, integrations, etc. I can do everything I want in FM with variables. I don't see what json adds. Maybe it's a server thing, of which I have zero knowledge. That's where you guys know the territory.

Exactly what I need! Where can I find version 3?

Another video from a member of the no comments club.

Portege Bay, one of the many high quality Filemaker Pro Developers.

Wow, M Rocharde is one of the greatest minds in Filemaker Pro. Nice to see he's still in the game.

Geez, got deep pretty quickly.

Where are the templates by crisippolite that are commented and all nicely scripted?

Nice work. How does one get a spot on the opinion page?

Hey, Matt! Recently discovered your new(ish) site from one of the podcast appearances you had. I love getting into the weeds on this stuff. Do you prefer that comments/discussions happen her or on your blog?

it would be much more of a useful tool if it was able to express filemaker code in the native fm xml file format. there is a tool called fmpastebox which enables to get and set the filemaker internal clipboard. but „someone“ must do the work first and make some spreadsheet mapping of the filemaker scriptsteps and every option of it with the corresponding xml expression. that way it would be hilariously capable

I completely agree, Matt. The validation features are totally unnecessary.

We have a CRM, so we 'have to have ' the Dial Phone script step... and/or surely there is no other option to communicate with MS Word 2024 other than Send DDE Execute... and/or the ADA guys might yell, but the Speak script step is on my 'never' list. Surely we (still) need the button and popover tools, because no one would understand a button bar with only one button. I suspect that your series may be a fun trip down memory lane for many; however, other items- like Summary fields; some users may not we aware of the more-modern approaches. Many of us realize why/ how these tools came in to being, but the larger question is why doesn't FileMaker remove some of these from the 'default' interfaces and allow users to control via Settings, much like they do with the 'Advanced Tools'. This approach could be similar to what Abode has done with Photoshop. Yes, I know we have Custom Menus, but I'm thinking of the Status Bar, Inspector, and other interfaces.... Fun topic...

why's everyone so old though

Excellent discussion! Thank you.

What’s missing is what was the version of FileMaker and the host OS version. Also UUID needs to check uniqueness due to its implementation of pseudo randomness etc…

No. Consciousness requires quantum processes.

Moriarty continued to exist outside the holodeck because the crew were still inside the holodeck

No

When you use RAG, are you giving up the ability of the AI to generalize? About 15 years ago, I read that AI should generalise, and if you use too many parameters and do too much training, you get overfitting, which kills the ability to generalise. For anyone not familiar with overfitting, it's easier to explain in terms of statistics, and I've made up an example. Suppose you have data for the age and height of every child in a school, you could fit a straight line to the data, and use the coefficients to look up the height you'd expect for a child at any age in the range. If the rate of getting taller varies with age, you could get a more accurate estimate by fitting a suitable curve that takes more than two parameters. If you fit too many parameters, the result is less useful. If you use more parameters than there are kids in the school, you're likely to get a wiggly curve that passes through every data point, corresponding to perfect recall of the data but with no ability to generalise. It would get like finding the child with the closest age, and that child might be unusually tall or short.

47:33 For some wild stuff don't stop at asking an LLM if it's a "database" , ask it to BE a relational database > Can you act like a relational database? then asking it > What commands do you have? vs > What database commands do you have? etc There's yt video 'mHgsnMlafwU' by @embracethered about this as well.

For a strained analogy if it's a "database" it's a randomly encrypted stochastic-storage database with no one having the decryption keys to the source fields. "Database" by itself implies structured deterministic retrieval of an original source. The entire LLM industry's core legal foundation of course is that original sources are used for training and are not stored, NOT retrievable. So any talk of a 'database' analogy for LLMS becoming a technical truth, even if proved as true, will always get pushback. So there's gonna be some illegitimacy|legitimacy to such an analogy|industry until some math geniuses prove retrievability is ALWAYS impossible AND that any thing seeming like an original source is an illusion of pure randomness; or deterministic retrieval is possible without having to store any random variables. Of coures they still have to communicate either outcome in terms that get talked about in conceptually simple ways. Either that or the law kicks in to say the randomness is/isn't good enough a defense. A fun thing for this is simple prompts like the following which seems likely to always provide an expected response seeming to show "original" information , especially when you reset the session per ask, but in a single session is completely random as you repeat the prompt multiple times without resetting (at least for chatgpt3.5): > finish the sentence: the quick brown fox

Great flow. Clear and informative. Subscribed

A good example of training being frozen in time - ask a model lile Gemini when the next total solar eclipse visible in the contiguous US will be. It's training stopped last year so the most probable completions point to April 8, 2024. If you let it know the response is incorrect, it will not infer that it needs to produce a result in the future but will answer with the eclipse prior to April's because that is the next most probable completion. Only if literally told the date of the next eclipse will it respond accurately...for that user/session.

I feel like a more accurate description is that it's a frozen database summarizer. It is truly incapable of putting two and two together on it's own. It cannot predict undocumented outcomes. Despite knowing infinitely more than any single human, it's unable to provide us with answers that encompass all that knowledge into better solutions. It can only feed us the very same conclusions that the more reliable humans have already come to. It feels like the real revolution will be when compute power is great enough for the training itself to be occuring in real time during every single conversation. Seems very far away!

It was great to chat! I'd love to hear people's feedback or questions.

I am SO appreciative you two are doing this podcast and immediately signed up when i saw it! One thing that would be amazing is bridging the gap between using a vector database in filemaker. or if there's someone that's already done it sharing the link, channel or podcast. I very much understand how vectors work and understand it's a huge set of numbers on given combinations of sentences and understand how to use api's and understand how to use OpenAI's chat api but i don't know where to go to get/put/collect/create the vector databases that i would then use against say search terms like you talk about. Any help is very much appreciated as i want to do this very badly!

I'm really enjoying these talks Matt and Chris. Being able to hear about your real world use cases, and thoughts on future integrations, gives me confidence my 25+ years of FileMaker will not yet be supplanted by AI. I'm more confident that, at least for the next few years, that AI will be a useful developer tool, extending my knowledge, and capabilities for my clients. Thank You. PS: The idea by Ernest of an "add this to AI knowledge base" button is a terrific one, lets hope Claris is watching and listening.

I’ve used UUIDNumber with the same assumption that it was faster. But I never did any benchmarking. This is early helpful and interesting. Thanks Matt.

Very interesting and informative! Thanks for posting this, Matt and Chris.👍

I did enjoy your content, as I'm working towards my certification on the Claris website and they don't have many visual experiences on their trainings, I finally understood relationships! Thanks for sharing it!

It is very strange do not see video! What’s has been happened?

You’re kind of all over the place, preparation is key when you what to convey information to novice users

Many great takeaways from this video. Thank you Matt!

Hi Matt- Neat vid. Have you seen any examples of someone connecting an FM database to GPT in order to interact with the data to provide answers to queries?

That's flippin' sweet. I much prefer hard data for creating best practices. Thanks for doing the work.

Great idea to put those two under stress! Very useful insights!

Thanks for that. The general consensus has usually been that the numeric approach is faster, but that's always been anecdotal. So it's nice to see a more rigorous approach to confirming or debunking that. Given the problems involved with numeric UUIDs in FileMaker when querying though OData, I'm not seeing any reasons why I would want to use numeric UUIDs in place of text.

I will have to see if it will write a custom function. That would be very handy. Anyone tried it?

How can get the sheet you used in this example ?

First thumbs up )

Fascinating subjetc. Thanks for your video. Are there any shortcuts to allow directly pasting the code in a script, rather than having to recreate line by line using the Script GUI? TY

Is it possible to actually copy/paste this code into the FileMaker Script editor?

Fascinating, Matt. It's all in the prompts! Thanks for posting.

Great job Matt. Loved it.

Terrific, Matt.

holy hell guys you might as well have remembered the joke 1 week later

I love Appalachian Spring and these closing moments are it's most sublime. Really enjoyed this. 👍❤