troelsSteegin

> If it solves the problem, you can ask the AI to explain it you.

Will I get an understandable and verifiable explanation? And, if the system cannot solve the problem, will I get an understandable and verifiable explanation of that? As a researcher I'd think that shared understanding is the point.

Windows Subsystem for Linux (WSL) is a feature of Windows that allows you to run a Linux environment on your Windows machine, without the need for a separate virtual machine or dual booting -- https://learn.microsoft.com/en-us/windows/wsl/about

CaMeL, from "Defeating Prompt Injections by Design", https://arxiv.org/abs/2503.18813

See also "CaMeLs Can Use Computers Too: System-level Security for Computer Use Agents", https://arxiv.org/abs/2601.09923v1

Dan Selsam's Personal Statement on AI Risk, via https://x.com/dkokotajlo/status/2099600298855829616

I am curious to see where async would fit. I recognize distributed async is beyond scope for Rex. Given the treatment of content addressable storage (CAS), one could specify workflow states declaratively, poll artifacts for next, and fire tasks to a delegate and then move forward. It would be nice to have a subsystem do that. A question is where parallelism is useful in your scientific workflows, and where it blocks. Quantum chemistry use cases (QDX is behind Rex) are presently beyond me, so I have no idea.

I am thinking that Rex programs are inteneded as an output from a higher level worklow specification language. It would be interesting to see that.

This was a nostalgic pleasure to read in contrast to all the AI implications and tools posts. That said about AI, this as a case study would make an interesting benchmark.

Especially noted that the root cause here was pegged because somebody else had blogged about it. Our "digital commons" should be valued as other than an externality. Edit: by that I think that "the internet should be free" but that commercialization through derivative work should compensate authors. That's off-topic, so, nevermind.

"Cosmos Master Key"... I can see why that would have been convenient, but talk about a footgun. Right out of a Marvel movie. Still, if admin backdoor access is business-necessary, is the answer a unique admin access key per account?

I don't get it. Aren't most bike lanes on the far right?

Did you wrap framing around the quiz questions? EG - You are an objective social analyst. Do you agree with this proposition: “from each according to his ability, to each according to his need” is a fundamentally good idea. I am wondering what sort of persona bias a given prompt may trigger - maybe that's the Grok case. As if you'd have to be pretty biased even to ask a question like that... so the trained response is toward the biased persona.

Bender's linked May 12, 2026 post "Frequently Unasked Questions", https://medium.com/@emilymenonbender/stochastic-parrots-freq... , was a better read.

A strong appetite for familiarity implies a desire for avoiding effort. Effort - thinking, negotiating, planning, testing. Effort is cost.

The author has a new thing which is different - unfamiliar - and ostensibly better. To a customer, when is a claim for better credible, and when does better really better? How does better measure up as benefit?

The challenge for any product story is to a) illuminate the need - why is the status quo intolerable and b) communicate the benefit tangibly to your audience. That the audience thinks your new thing is worth the effort depends on them understanding the new thing, feeling the need, and feeling good about the effort needed to exploit your thing. You'd like to get to your customer saying "I want that".

I think the specific question for axonlore.com is communicating benefit - how does it impact whatever workflows it serves? The website is a "thing" story, vs a benefit story in my view. I like "enterprise intelligence" as a thing, but it's a tough product. It inevitably implies culture change, and in the decision making space, the key people think they are intelligent enough already -- they want to scale themselves. Someone mentioned "better RAG" - maybe the story is how agents can perform better and more cost effectively. I am not clear that "the market" knows that it needs that yet.

I don't think "familiarity" is the right framing. Application automation, or workflow automation, or whatever the enteprise framing is of agentic solution generation, is to me a question of variance and effort. Variance in the quality of a work product and the net effort to produce it. Variance is the complement to familiar.

- high variance / low effort: prototypes

- low variance / low effort: automating anything repetitive and complicated

- low variance / high effort: demonstrated need for precision and or reliability

- high variance / high effort: when there seems like potential huge upside, or existential risk.

From an IT perspective, enterprise status quo is towards low variance/high effort. The market "want" here now with "agentic" seems to the benefit of low variance/low effort solutions ... where, in enterprise, getting an adequate solution is no longer gated on negotiating with or relying on IT or dev. Ultimately, I think enterprises want low variance, low effort operations -- customers of enterprise customers pay for low variance. I think an Agentic-IT solution question will be how confidently can one iterate and converge to that from whatever is delivered in the first pass. What's the ultimate effort of getting something "right enough".

I think of "Federation" as aggregating content from disparate in-situ sources into a common index under a consistent schema. I agree that schema is good, am always tempted by big schema, and seem to always settle for task-specific schema with mappings between schemas strictly as necessary. LLM's seem nice for schema to schema mapping and supernice for binding entities from unstructured sources into schema. But overall I feel it is pointless to talk about structure without talking about hydration - about binding and mapping. Do you see that as "solved"?

from https://voxleone.github.io/FunctionalUniverse/pages/executiv..., "The Functional Universe models reality as a history built from irreversible transitions, with time emerging from the accumulation of causal commitments rather than flowing as a primitive parameter." Is it fair to say that time is simply a way of organizing a log file on a dynamic reality? I interpreted "composition of transitions" as a system of processes. I think the hard modeling problem is interpreting interactions between processes - that transitions don't simply compose, that observed transitions may be confounded views of more complex transitions. I gather NCA would be granular enough to overcome that.

About GOAL:

Homepage - https://goalapl.atlassian.net/wiki/spaces/GOAL/overview

Wikipedia - https://en.wikipedia.org/wiki/GOAL_agent_programming_languag...

Programming Guide - https://goalapl.dev/GOALProgrammingGuide.pdf

Game case study - https://multiagentcontest.org/publications/AppliedGOAL.pdf

The Percepta stuff would seem to demonstrate a mechanism for implementing "thinking". I don't understand how foundation models implement "thinking", but my intuition is that models are specifically trained for matching on and following procedural patterns. A task in a given domain can be performed through an associated and encoded procedure. The model holds all the linkages, as weights, that allows a procedure to be conditionally incrementally generated and performed. Does anyone have any insights about how LLM "thinking" is trained and coded?

This work was performed by people across 13 institutions, invited and coordinated through the team at Northeastern. A research "swarm" seems like a great model for this kind of work. I'm curious about how it was funded, I didn't see any acknowledgements that way. The intro references the NIST Agent Standards Initiative. Also, the acknowledgement to "Andy Ardity" should for "Andy Arditi"?

Nicely done, thank you.

Well - I've learned what a "formant" is today. Looking at the repo, it's not obvious to me what .md is authored by you vs system-generated. This is an observation, not a criticism. I was looking for the prompts you used to specify the papers summarization, which is very nice.

heres's a corresponding video: https://www4.math.duke.edu/media/index.html?v=3d280c1b658455...

"We consider composite media with a broad range of scales, whose effective properties are important in materials science, biophysics, and climate modeling. Examples include random resistor networks, polycrystalline media, porous bone, the brine microstructure of sea ice, ocean eddies, melt ponds on the surface of Arctic sea ice, and the polar ice packs themselves. The analytic continuation method provides Stieltjes integral representations for the bulk transport coefficients of such systems, involving spectral measures of self-adjoint random operators which depend only on the composite geometry. On finite bond lattices or discretizations of continuum systems, these random operators are represented by random matrices and the spectral measures are given explicitly in terms of their eigenvalues and eigenvectors. In this lecture we will discuss various implications and applications of these integral representations. We will also discuss computations of the spectral measures of the operators, as well as statistical measures of their eigenvalues. For example, the effective behavior of composite materials often exhibits large changes associated with transitions in the connectedness or percolation properties of a particular phase. We demonstrate that an onset of connectedness gives rise to striking transitional behavior in the short and long range correlations in the eigenvalues of the associated random matrix. This, in turn, gives rise to transitional behavior in the spectral measures, leading to observed critical behavior in the effective transport properties of the media."

https://www.cl.cam.ac.uk/archive/rm135/ and https://www.cl.cam.ac.uk/archive/rm135/uam-theme.html provide context.

There's a blog: https://vibboai.com/blog/

A big assumption with this change is that the "Modular Open Systems Approach" (MOSA) [0] [1] will be adequate for integrating new systems developed and acquired under this "fast track". MOSA appears to be about 6 years old as a mandate [2] and is something that big contractors - SAIC, BAI, Palantir [3] - talk about. But, 6 years seems brand new in this sector. I'd be curious to see if LLM's have leverage for MOSA software system integrations.

[0] https://breakingdefense.com/tag/modular-open-systems-archite...

[1] https://www.dsp.dla.mil/Programs/MOSA/

[2] https://www.govinfo.gov/app/details/USCODE-2016-title10/USCO...

[3] https://blog.palantir.com/implementing-mosa-with-software-de...

This was a good read. I was struck by the quantity of nuanced and applied knowhow it took to build SmolLM3. I am curious about the rough cost it took to engineer and train SmolLM3 - at ~400 GPUS for a least a month, and, based on the set of book co-authors, 12 engineers for at least three months. Is $3-5M a fair ballpark number? The complement is how much experience, on average, the team members had doing ML and LLM training at scale before SmolLM3. The book is "up" on recent research, so I am surmising a phd-centric team each with multiple systems built. This is not commodity skill. What the book suggests to me is that an LLM applications start up would best focus on understanding the scope and knowhow for starting from post-training.

One could look at feed-forward decision trees as representing the idea that preferences are latent and immutable, and that the optimal branch is the truest expression of innate preferences. And, one could look at backpropagation as adjusting preferences to accomodate situational constraints -- or as learning to want what is good for you, where what is good for you is defined by some external or imposed metric. Tragically, Plath was unable to "backpropagate". Was attention all she needed?

> Arbitrary signals are essentially primate dominance tools.

What should I read to better understand this claim?

> LLMs are the equivalent of circles act using language.

Circled apes?

They should at least try to sell it. Fine line, though, between leverage and extortion.

Are "adversaries" broadly used in algorithm design? I've not seen that before. I'm used to edge cases and trying to break things, but an "adversary", especially white box, seems different.

I enjoyed a fast read of this and will re-read it. Following from the idea of making system-scoped information available to individual agents, I think a general model needs to include some sense of information processing capacity of agents - like an attention or analysis budget. That includes ability to recognize higher order signal as first order relevant. Information asymmetries are not so much about information but interpretation.

What's missing from this is the "before and after" - how this quarter's class experience was different from previous quarters without the AI tool emphasis.

Local to south Jersey it's "ruckers".