Corpus Relica

Semantic Modeling for Next Generation Cognitive Applications

[Amelia Wattenberger 🪷 @Wattenberger]

I have a problem with signing up for too many newsletters (they're free!!).

going through them this morning, I tried exploring some in the semantic sphere, and it was really interesting.

I'm not sure how I feel about it yet, but it feels very different from reading. Nice to skim a bit and see if there's anything that pulls me to dig in more.

[Geoffrey Litt @geoffreylitt]

the real computer revolution may be happening

[𝐺𝑖𝑠𝑒𝑙𝑙𝑒 @wildspiritlibre]

The Ouroboros

[Kevin Twohy @kevintwohy]

This book is over 30 years old. It articulates a future where software and technology are accepted as safe and obviously beneficial conduits for creativity, play, and learning.

None of it has come to pass because we failed to create suitable form factors and ergonomics. Complete failure of imagination on the part of adults.

[Marc Christophe @fusupo]

AI beat the world champion at Go. It still can't tell you which supplier is unreliable or which order is late.

Those aren't spatial facts. They're relational facts. And they're the world model that actually matters.

https://t.co/jy0e8pZSAO

[Marc Christophe @fusupo]

Knowledge and reasoning aren't fundamentally text. The "world model" as visual/physical predictor risks recreating the same integration problem at a different layer. Reality isn't visual or physical, it's an information space of interrelated concepts.
https://t.co/st9xOifKo0

[Corpus Relica @CorpusRelica]

@xbh_artist this is what it feels like to be an insect

[Corpus Relica @CorpusRelica]

@GaryMarcus The world is also not just physical and visually spatial. Or to put it another way, just as language can be seen to be the surface of knowledge, so too what is perceived visually can be seen to be only the surface of reality.

[Marc Christophe @fusupo]

A book has weight. The novel does not.
We build oceans of cable and silicon to move things that have no mass. Worth thinking about.
https://t.co/mDEY87TMxJ

[Andrej Karpathy @karpathy]

LLM Knowledge Bases

Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).

Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.

Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.

Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.

Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.

Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.

TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

[Marc Christophe @fusupo]

The AI industry made a foundational bet: structured representation is over. Model weights hold the knowledge. Context windows hold the state. Just prompt it, bro.
That bet is losing. Here's what comes next. 🧵

[Marc Christophe @fusupo]

Structured representation isn't over. It never was. The "AI bubble" is the reluctance to admit the walkback from that initial overestimation.

Chat Is Not Where It's At
https://t.co/SwEcqc19ZC

[{ I N P U T } @lt0gt]

How it feels to chew 5 claude

[Jeff Clune @jeffclune]

Can AI agents design better memory mechanisms for themselves?
Introducing Learning to Continually Learn via Meta-learning Memory Designs. A meta agent automatically designs memory mechanisms, including what info to store, how to retrieve it, and how to update it, enabling agentic systems to continually learn across diverse domains. Led by @yimingxiong_ with @shengranhu 🧵👇 1/

[maxim zhestkov @maximzhestkov]

I believe that the interfaces of the future will disappear. Instead of buttons and menus, we will interact with objects. The process itself becomes the feedback.

Building software is always a complex craft, but spatial computing is a different beast entirely. It demands a fusion of visual art, logic, and spatial UI where every interaction happens in the air around you.

With OPERATOR, built for Apple Vision Pro, we explored this by turning music into a sculpture. The music track isn't a timeline anymore; it’s a 3D object you can walk around, touch, and modify from any angle.

[Marc Christophe @fusupo]

Escapement
https://t.co/y0diHDiMKT

outlining my claude-code workflow and philosophy

[Marc Christophe @fusupo]

What even is a computer?
https://t.co/4vVYKt7MzP

[Marc Christophe @fusupo]

If 2026 is the year AI changes everything, shouldn't that blow back into the UI?
Chat works for questions. Alone, it's inadequate for the breadth of thinking we want to do with these systems.
https://t.co/xjNJMFJnEl

[Alex Veremeyenko @alex_verem]

This paper from Harvard and MIT quietly answers the most important AI question nobody benchmarks properly:

Can LLMs actually discover science, or are they just good at talking about it?

The paper is called “Evaluating Large Language Models in Scientific Discovery”, and instead of asking models trivia questions, it tests something much harder:

Can models form hypotheses, design experiments, interpret results, and update beliefs like real scientists?

Here’s what the authors did differently 👇

• They evaluate LLMs across the full discovery loop hypothesis → experiment → observation → revision
• Tasks span biology, chemistry, and physics, not toy puzzles
• Models must work with incomplete data, noisy results, and false leads
• Success is measured by scientific progress, not fluency or confidence

What they found is sobering.

LLMs are decent at suggesting hypotheses, but brittle at everything that follows.

✓ They overfit to surface patterns
✓ They struggle to abandon bad hypotheses even when evidence contradicts them
✓ They confuse correlation for causation
✓ They hallucinate explanations when experiments fail
✓ They optimize for plausibility, not truth

Most striking result:

`High benchmark scores do not correlate with scientific discovery ability.`

Some top models that dominate standard reasoning tests completely fail when forced to run iterative experiments and update theories.

Why this matters:

Real science is not one-shot reasoning.

It’s feedback, failure, revision, and restraint.

LLMs today:

• Talk like scientists
• Write like scientists
• But don’t think like scientists yet

The paper’s core takeaway:

Scientific intelligence is not language intelligence.

It requires memory, hypothesis tracking, causal reasoning, and the ability to say “I was wrong.”

Until models can reliably do that, claims about “AI scientists” are mostly premature.

This paper doesn’t hype AI. It defines the gap we still need to close.

And that’s exactly why it’s important.

[DailyPapers @HuggingPapers]

Memory in the Age of AI Agents: A Survey

Delves into the rapidly evolving field of agent memory, providing a unified taxonomy through the lenses of forms, functions, and dynamics. This survey offers a critical conceptual foundation for designing future agentic intelligence.