There is a reason we call them language models. It is the kind of sentence that sounds obvious until you sit with it and look around your digital environments. Large Language Models. Not large knowledge models or large reasoning models or large world models. Language is the substrate and the bridge, the surface with which we interact with systems and interfaces. Strip the language away and what remains is a tensor of weights with no bearing on the real world.
The era of anonymous calls may well be coming to an end. On 30 April, the Federal Communications Commission unanimously approved a proposal requiring telecoms operators to verify their customers’ identities before activating the service.
A government-issued ID, a physical address, a legal name and existing telephone numbers would all be required. The stated aim is to put an end to automated calls
L’ère du numéro de téléphone anonyme pourrait bien prendre fin. Le 30 avril, la Commission fédérale des communications a approuvé à l’unanimité une proposition obligeant les fournisseurs de télécommunications à vérifier l’identité des clients’ avant d’activer le service.
Une pièce d’identité délivrée par le gouvernement, une adresse physique, un nom légal et des numéros de téléphone existants seraient tous inclus. L’objectif déclaré est de mettre fin aux appels automatisés
The date and location are set: On May 21, Sana AI Summit 2026 takes place at The New York Public Library. We'd love for you to join us.
Every year, Sana AI Summit brings together the people asking the hardest questions about where AI is heading. This year's lineup features economist Tyler Cowen, Lovable co-founder Anton Osika, and author Benjamín Labatut — and previous years brought Jensen Huang, Geoffrey Hinton, and Garry Kasparov to the same stage.
Seats are extremely limited. All registered guests will receive recordings
Eliminating the ambiguity of language is essential when we want to use words as mathematical symbols to drive agreement and organizational alignment. (Before we all agree on how to handle inactive accounts, we need to agree on what we mean by inactive accounts.)
But I recall what a career coach once said to me - the thing that divides middle management and top leadership is very often the "ability (or inability) to deal with ambiguity." In many situations ambiguity of meaning is the lubricant that allows decisions to be made, even when they are not the best decision. It allows disagreement to hide behind ambiguous terms. (We all agree that inactive accounts should be removed, but if never agree to what that terms mean, we get to do our own things while pretending to agree.)
I wonder if ontology development is often hard because people don't want to come to consensus with others on terms, or even resist being locked down on their own definitions. (my personal use of the term inactive accounts changes from context to context.)
If we want to speak to our LLM's in knowledge graphs, we may be forced to give up some of the comfort of ambiguity.
This demo took under three minutes, using OpenAI’s Codex or Anthropic’s Cowork, leveraging the fact that these AI agents now relegate browsers to highly controllable features within the agent experience. The steps were as follows:
1. Ask the AI agent to generate notes in both HTML and RDF from your article using a template comprising a FAQ, glossary, and how-to (where applicable).
2. Copy the HTML and RDF documents to a remote folder in my Virtuoso-based public data space.
3. Rinse and repeat the next time I encounter an article of interest.
Benefit?
A progressively enhanced knowledge graph component within my data space 🙂
I can add support for RDFa, just like any other RDF notation. That’s the beauty of these LLMs—they’re good at language-related tasks, but not so good at deterministic tasks better handled by databases, knowledge bases, filesystems, etc.
On a side note, just last week I had an engaging convo with one of my friends whose book on Wittgenstein's math philosophy will be published later this year.
Jessica, this is such an exciting piece for me. I struggle so hard from my side of the fence to articulate this—I know it exists but because I don’t work in the computational side of linguistics, I don’t have all the reference points.
I’m focusing on Frame Theory at the moment and you reminded me of Filmore’s FrameNet which in of itself is fascinating as a global database.
As someone who’s needed to read this from someone else, thank you 🙏🏽
Another fantastic post Jessica. Thank you. Always learning something from you. Cheers.
The era of anonymous calls may well be coming to an end. On 30 April, the Federal Communications Commission unanimously approved a proposal requiring telecoms operators to verify their customers’ identities before activating the service.
A government-issued ID, a physical address, a legal name and existing telephone numbers would all be required. The stated aim is to put an end to automated calls
Hi 👋 😉 🇫🇷
L’ère du numéro de téléphone anonyme pourrait bien prendre fin. Le 30 avril, la Commission fédérale des communications a approuvé à l’unanimité une proposition obligeant les fournisseurs de télécommunications à vérifier l’identité des clients’ avant d’activer le service.
Une pièce d’identité délivrée par le gouvernement, une adresse physique, un nom légal et des numéros de téléphone existants seraient tous inclus. L’objectif déclaré est de mettre fin aux appels automatisés
The date and location are set: On May 21, Sana AI Summit 2026 takes place at The New York Public Library. We'd love for you to join us.
Every year, Sana AI Summit brings together the people asking the hardest questions about where AI is heading. This year's lineup features economist Tyler Cowen, Lovable co-founder Anton Osika, and author Benjamín Labatut — and previous years brought Jensen Huang, Geoffrey Hinton, and Garry Kasparov to the same stage.
Seats are extremely limited. All registered guests will receive recordings
https://news.ycombinator.com/item?id=48023079
Eliminating the ambiguity of language is essential when we want to use words as mathematical symbols to drive agreement and organizational alignment. (Before we all agree on how to handle inactive accounts, we need to agree on what we mean by inactive accounts.)
But I recall what a career coach once said to me - the thing that divides middle management and top leadership is very often the "ability (or inability) to deal with ambiguity." In many situations ambiguity of meaning is the lubricant that allows decisions to be made, even when they are not the best decision. It allows disagreement to hide behind ambiguous terms. (We all agree that inactive accounts should be removed, but if never agree to what that terms mean, we get to do our own things while pretending to agree.)
I wonder if ontology development is often hard because people don't want to come to consensus with others on terms, or even resist being locked down on their own definitions. (my personal use of the term inactive accounts changes from context to context.)
If we want to speak to our LLM's in knowledge graphs, we may be forced to give up some of the comfort of ambiguity.
...but is there also a danger that the humans become part of the Machine? :) Thinking of Bradbury here haha.
ZH
haha…but yes
Yep!
Language is mankind's greatest innovation :)
Isn't it though? I have experienced all too many times ontologies designed and implemented with no consideration for language.
The connection isn’t immediately obvious—at least not yet. It’s similar to the powerful relationship between RDF and computational linguistics.
What’s needed is continued education through articles like this, along with supporting collateral.
For example, here are notes from this post expressed in RDF using terms defined in various ontologies:
[1] https://linkeddata.uriburner.com/DAV/demos//daas/language-is-the-bridge-infographic.html
[2] https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fjessicatalisman.substack.com%2Fp%2Flanguage-is-the-bridge%23Post&distinct=1
This demo took under three minutes, using OpenAI’s Codex or Anthropic’s Cowork, leveraging the fact that these AI agents now relegate browsers to highly controllable features within the agent experience. The steps were as follows:
1. Ask the AI agent to generate notes in both HTML and RDF from your article using a template comprising a FAQ, glossary, and how-to (where applicable).
2. Copy the HTML and RDF documents to a remote folder in my Virtuoso-based public data space.
3. Rinse and repeat the next time I encounter an article of interest.
Benefit?
A progressively enhanced knowledge graph component within my data space 🙂
I have been wondering and potentially super rudimentary, but do you leverage RDFa?
I have been scratching my head about the abandonment of this spec and figured you may know a thing or two 😉
https://www.w3.org/2010/02/rdfa/sources/rdfa-primer/Overview-src.html
I can add support for RDFa, just like any other RDF notation. That’s the beauty of these LLMs—they’re good at language-related tasks, but not so good at deterministic tasks better handled by databases, knowledge bases, filesystems, etc.
Ah, these French people, they don't want to learn, they're so stubborn.
I thoroughly enjoyed this piece, Jess, thank you!
On a side note, just last week I had an engaging convo with one of my friends whose book on Wittgenstein's math philosophy will be published later this year.
Thanks, Ramona. I’ll look forward to your friend’s book!
Language is indeed the bridge.
But a bridge to what?
If language is not grounded in the event that generates meaning,
then even the most refined ontology risks becoming fluent emptiness.
Perhaps the real problem is not that models lack language,
but that we have not yet understood how meaning itself emerges before it stabilizes into language.
Jessica, this is such an exciting piece for me. I struggle so hard from my side of the fence to articulate this—I know it exists but because I don’t work in the computational side of linguistics, I don’t have all the reference points.
I’m focusing on Frame Theory at the moment and you reminded me of Filmore’s FrameNet which in of itself is fascinating as a global database.
As someone who’s needed to read this from someone else, thank you 🙏🏽