Forum Posts Feed
Tinderbox goes AI
< Next Topic | Back to topic list | Previous Topic >
Posted by Paul Korm
Aug 12, 2025 at 09:22 PM
Looks like Eastgate is working on a new version of Tinderbox that uses the MCP framework to integrate Claude desktop to a Tinderbox document. Might be interesting and useful. Might be an attention sink. One day we’ll have brave developers who assure us there is no reason to use their product with LLMs.
Posted by MadaboutDana
Aug 13, 2025 at 02:36 PM
Indeed. It’s worth reading this article for tech insights into exactly how rushed and unprofessional MCP is: https://julsimon.medium.com/why-mcps-disregard-for-40-years-of-rpc-best-practices-will-burn-enterprises-8ef85ce5bc9b
Paul Korm wrote:
Looks like Eastgate is working on a new version of Tinderbox that uses
>the MCP framework to integrate Claude desktop to a Tinderbox document.
>Might be interesting and useful. Might be an attention sink. One day
>we’ll have brave developers who assure us there is no reason to use
>their product with LLMs.
Posted by eastgate
Aug 16, 2025 at 08:55 PM
I think I have a decent record of directing my attention! In any case, MCP support is (a) not terribly difficult, and (b) absolutely fascinating. I’m writing a series at https://markBernstein.org/ about (b).
Julien Simon’s piece conflates all the many scenarios for cooperation. Yes: CORBA was fascinating! Have you opened anything with OpenDoc lately? Sure SOAP was great. Absolutely, banking apps need lots of care.
But the lesson of the Web — really the lesson of the 21st century — is that you engineer everything as if it were a bank, you’ll easily be out-maneuvered and outrun. I say this with some reluctance: I was on the program committee that rejected the initial research paper on WWW because its treatment of dangling links — 404s — was superficial and ignored lots of good prior work. Berners-Lee and Caillau were right, though: the simplicity of the protocol really matters for adoption.
I write a tool for analyzing and visualizing notes, and I’m writing a book about reimagining the intellectual position of Computing. I’m a gu, with lots of questions. Claude is astonishingly good at locating good sources. Not perfect, but no research assistant is. It is extraordinarily well read. It can read a crash log. It can read a man page. It can find the best reference for Nero’s rotating dining room, which is something I actually needed for the book the other day.
I think that’s promise enough to merit a few days of development work.
MadaboutDana wrote:
Indeed. It’s worth reading this article for tech insights into
>exactly how rushed and unprofessional MCP is:
>https://julsimon.medium.com/why-mcps-disregard-for-40-years-of-rpc-best-practices-will-burn-enterprises-8ef85ce5bc9b
>
Posted by Lucas
Aug 17, 2025 at 03:17 PM
This is fascinating. Just a quick side note: I think Paul was referring to the attention of users rather than the attention of developers, and Claude seems to agree with me :-)
(My own view is that AI integration is both potentially valuable also probably important for staying relevant and competitive.)
Posted by Amontillado
Aug 17, 2025 at 09:00 PM
AI is pretty amazing, particularly if your definition of “amazing” includes a tinge of terror.
Or at least humor.
Yesterday, I wrote a letter that weighed 0.995 ounces by my scales. It could easily read 1.005 ounces on a different scale.
The resolution of postal measurement was of critical interest, so I asked Google if I should put more postage on a 0.99 ounce letter. Would my measurement of 0.99 fall within tolerance if the post office saw a different weight?
Absolutely not, the Google AI bot told me. A first class stamp covers one ounce. Since 0.99 ounces is far more than one ounce, I’d have to lick more stamps.
So I asked the question a different way, got a different answer, but the conclusion was the same. Since 0.99 ounces is so much more than one ounce I would have to use more postage.
In that experience I think I saw the downfall of mankind.
I added postage. Not because I thought 0.99 ounces was more than one but because I didn’t want to risk nondelivery.
Sensible, I thought, until I realized the AI’s silly argument prevailed. Whether or not I believed 0.99 > 1, my actions paralleled a ludicrous conclusion.
If that’s not the end of the world, it’s at least the beginning of a lot of post AI political careers.
We’re sunk.