Both were better in terms of user ergonomics and also much faster that what we used before, black, mypy, pylint etc.
IDK it Typescript is too generic but TS made me think of javascript as a proper language and not something just to do silly little animations on websites for. Types unlock proper data modeling and application code to me.
Figma and Figjam are really great tools for design and planning. I didn't like Miro's pricing model, and stuff we used before like Sketch wasn't as good.
When I need to create something like a powerpoint or whatever I use claude code and invoke a claude skill that knows how to do it. Why would I use claude cowork instead of that?
Thanks for the direct push. Let me ground those statements in the framework of the paper:
1. On "eroding human agency in a black box":
I am referring to "Agency Misattribution". When Generative AI transitions from a passive tool to an active agent, it silently corrects and optimizes human input without explicit consent. The evidence is observable in the psychological shift where users internalize system-mediated successes as personal mastery. For example, when an LLM silently polishes a draft, the writer claims authorship over nuances they did not actually conceive.
2. On "healthy coexistence":
In this paper, this is defined as "Seamful Agency". It is a state where the human can quantify the "D" (Discrepancy) between their raw intent and the system's output. Coexistence is "healthy" only when the locus of judgment remains visible at the moment of intervention.
For a more rigorous definition of JTP and the underlying problem of "silent delegation," I highly recommend reading Chapter 1 of the white paper.
Does this technical framing of "agency as a measurable gap" make more sense to you?
Use Hammerspoon [0][1], it comes with a lot of macOS integrations out of the box and you write Lua, which takes zero effort to pick up and use. For me a big benefit is that you don't need to touch Xcode at all.
Speaking of the macOS menu bar, is there some way to make it overflow into a hidden area behind an "expand" button that would appear when there is no more room, like the notification area in Windows does?
The macOS UI decision of "just pretend that whatever doesn't fit to the right of the notch doesn't exist" is baffling.
I've seen a few apps that claim to do that, but it's always done in some really hacky way (such as needing screen recording permissions), and the behavior is never that of simple overflow handling. Instead they have "always hidden" sections and things like that, which is not what I want.
I used CC just yesterday to build a native MacOS menu bar app by using plan mode (opus) until there was alignment then to edit mode for the build (I use Zed for the prompting).
CC walked me through the needed Xcode project setup and handled all of the code there after.
I’m sure something more complex would be more challenging but I was happy with a two-shot result for this native menu bar app.
FTs are actually very reasonable, in the sense that they are a easy to reason about conceptually and in practice.
There's another title referenced in that link which is equally asinine: "Eugene Wigner's original discussion, "The Unreasonable Effectiveness of Mathematics in the Natural Sciences". "
Like, wtf?
Mathematics is the language of science, science would not compound or be explainable, communicable, or model-able in code without mathematics.
It's actually both plainly obvious for mathematics then to be extremely effective (which it is) and also be evidently reasonable as to why, ergo it is not unreasonably effective.
Also the slides are just FTs 101 the same material as in any basic course.
Hi, original presenter here :) The beginning is FTs 101. The end gets more application-centric around OFDM and is why it feels 'unreasonably effective' to me. If it feels obvious, there's a couple of slides at the end that are food for thought jumping off points. And if that's obvious to you too, let's collab on building an open source LTE modem!
If one wants to contribute to an open-source LTE modem, the best place to start may be OpenLTE: https://en.wikipedia.org/wiki/OpenLTE The core of any LTE modem is software, even if it is written for DSPs or other low-level software.
> FTs are actually very reasonable, in the sense that they are a easy to reason about conceptually and in practice.
ok but it's not the FTs that are unreasonable, it's the effectiveness
I think we all understand at this point that "unreasonable effectiveness" just means "surprisingly useful in ways we might not have immediately considered"
I find it hard to parse the middle of your post. Are you saying Wigner's article, which is what all the "unreasonable effectiveness" titles reference, is silly?
If that is what you are saying I suggest that you actually go back and read it. Or at least the Wiki article:
By means of contrast: I think it's clear that mathematics is, for example, not unreasonably effective in psychology. It's necessary and useful and effective at doing what it does, but not surprisingly so. Yet in the natural sciences it often has been. This is not a statement about mathematics but about the world.
(As Wittgenstein put it some decades earlier: "So too the fact that it can be described by Newtonian mechanics asserts nothing about the world; but this asserts something, namely, that it can be described in that particular way in which as a matter of fact it is described. The fact, too, that it can be described more simply by one system of mechanics than by another says something about the world.")
Yeah it's silly, I don't mean it in any mean spirited way.
> Wigner's first example is the law of gravitation formulated by Isaac Newton. Originally used to model freely falling bodies on the surface of the Earth, this law was extended based on what Wigner terms "very scanty observations"[3] to describe the motion of the planets, where it "has proved accurate beyond all reasonable expectations."
So despite 'very scant observations' they yielded a very effective model. Okay fine. But deciding they should be 'unreasonably' so is just a pithy turn of phrase.
Something can be effective, and can be unreasonably so if it's somehow unexpected, but I basically disagree that FTs or mathematics in general are unreasonably so since we have so much prior information to expect that these techniques actually are effective, almost obviously so.
I am not discussing the FT case. But as regards Wigner's article, the core thing he points out is that while we are used to the effectiveness of maths, centuries after Newton, there in fact is not any prior grounds to expect this effectiveness.
And no, this is unrelated to whether math is invented or discovered. If anything this is related to the extreme success of reductionism in physics.
As a general point of reflection: If an influential article by a smart person seems silly to you, it's good practice to entertain the question if you missed something, and to ask what others are seeing in it that you're missing.
So, biology and medicine are not sciences? Or are only sciences to the extent they can be mathematically described?
The scientific method and models are much more than math. Equating the reality with the math has let to myriad misconceptions, like vanishing cats.
And silly is good for a title -- descriptive and enticing -- to serve the purpose of eliciting the attention without which the content would be pointless.
They are still capable of being described with math, we are just not capable of doing the math, or probably better put is there is a diminishing return of doing the formalisation of those systems as our cognitive abilities are limited and couldn't reason about those models. It leaves them using very approximate models based on human language descriptions that can be reasoned about.
Which means, the language of some fields can’t be math.
However, I don’t think the original presenter was asserting those fields aren’t science, that’s an unreasonable interpretation. More so , ideally they would be use math as it is a language that would help prevent the silly argument “so, Y is not X?, or is Y only X provided Y is in the subset of X that excludes Z? “
(Even in Engineering, we hit this cognitive limit, and all sorts of silliness emerges about why things are or are not formalised)
It is likewise unreasonable to look down on any kind of world model from the past. Remember that you, in 2026, are benefitting from millions of aggregate improvements to a world model that you've absorbed passively through participation in society, and not through original thought. You have a different vantage point on many things as a result of the shoulders of giants you get to stand on.
I mean... this one's actually a pretty good paper, but we also had Linus Pauling pontificate on Vitamin C, so maybe we should cool it with the appeals to Nobel authority alone.
It's not easy to separate cause and effect from direct and strong correlations that we experience.
The job of a scientist is not to give up on a hunch with a flippant "correlation is not causation" but pursue such hunches to prove it this way or that (that is, prove it or disprove it). It's human to lean a certain way about what could be true.
You can make a router skill that describes how to use the other skills together. I'm experimenting with this now but my core problem is still How to Make Claude Code Skills Activate Reliably.
During testing today I asked a task I knew should have activated a skill and claude just did it without the skill instead.
My big gripe with skills is getting claude webapp and claude code to get them to invoke them in the right situations (often unexpected situations) without explicitly telling it directly to use skill x + skill y.
Ideally I would build a bunch of atomic skills that combine well and claude just uses them naturally when the situation arises.
Edit: I realise it might look kind of weird I posted this link in 2 comments on this thread - disclaimer I am not the author of that random blog post, just sharing what I found!
The amount of people holding strong opinions on LLMs who openly admit they have not tried the state of the art tools is so high on Hacker news right now, that it's refreshing to get actual updates from the tool's creators.
I read a comment yesterday that said something like "many people tried LLMs early on, it was kind of janky and so they gave up, thinking LLMs are bad". They were probably right at the time, but the tech _has_ improved since then, while those opinions have not changed much.
So, yes claude code and sonnet/opus 4.5 is another step change that you should try out. For $20/month you can run claude code in the terminal and regular claude on the web app.
uv solved packaging.
Both were better in terms of user ergonomics and also much faster that what we used before, black, mypy, pylint etc.
IDK it Typescript is too generic but TS made me think of javascript as a proper language and not something just to do silly little animations on websites for. Types unlock proper data modeling and application code to me.
Figma and Figjam are really great tools for design and planning. I didn't like Miro's pricing model, and stuff we used before like Sketch wasn't as good.
reply