pagekicker
pagekicker on Hacker News
No, it's not, unless you can argue that the bot would have thought of asking the same question I did, which is unlikely.
I asked Grok to visualize this:
https://grok.com/share/bGVnYWN5_463d51c8-d473-47d6-bb1f-6666...
*Caption for the two images:*
Artistic visualization of the universal low-parameter subspaces discovered in large neural networks (as described in “The Unreasonable Effectiveness of Low-Rank Subspaces,” arXiv:2512.05117).
The bright, sparse linear scaffold in the foreground represents the tiny handful of dominant principal directions (often ≤16 per layer) that capture almost all of the signal variance across hundreds of independently trained models. These directions form a flat, low-rank “skeleton” that is remarkably consistent across architectures, tasks, and random initializations.
The faint, diffuse cloud of connections fading into the dark background symbolizes the astronomically high-dimensional ambient parameter space (billions to trillions of dimensions), almost all of whose directions carry near-zero variance and can be discarded with negligible loss in performance. The sharp spectral decay creates a dramatic “elbow,” leaving trained networks effectively confined to this thin, shared, low-dimensional linear spine floating in an otherwise vast and mostly empty void.
like this idea. Add "tech trees": path dependence can be arbitrary. What if we kept going with vacuum tubes/no transistors?
Codexes Factory: algorithmic tools to create, operate, distribute, and market entire publishing imprints. This week I am launching my first imprint, Xynapse Traces, with 66 books in the Korean pilsa (筆寫) style. Later in October, Nimble Ultra, devoted to the history and practice of intelligence and espionage. Last week I built a giant collection of 575 imprints that are a shadow superset of the ~540 imprints operated by the Big Five publishing houses (Penguin Random House, the largest has ~300). Teeny weeny tip of the iceberg at NimbleBooks.com.
Pilsa is a form of journalling popular in Korea that should be popular everywhere.
AI Safety Event: DOGE is analyzing the entire Code of Federal Regulations this month with the goal of reducing it by 50% by Jan 20, 2026. [WaPo 7/26/25].
--This is an actual AI governance crisis that is happening today.--
Mitigation: Enable OSS analysis of CFR by all parties using same data with different goals.
https://github.com/fredzannarbor/altDOGE/tree/mainhttps://op...
First results:
As a proof of concept, ran altDOGE v0.2 against 485 regulations for the Community Living Administration using DOGE-like prompts, identified likely areas for cuts. Appears to work as intended.
To continue this mitigation effort, the following is needed:
-- policy participants with agency-specific expertise to carry out real validation.
-- LLM analysis experts
-- inference resources
I see no privacy advantage to working with Ollama, which can sell your data or have it subpoenaed just like anyone else.
Per the Washington Post, DOGE is running LLM analysis against the Code of Federal Regulations (5B tokens) *this month* (August) with the goal of generating revision texts that will eliminate 50% of US regulations. This is a governance cataclysm that is happening right now. My proposal is that there should be many teams analyzing the same data and making recommendations.
A lot of vibrant color, which is a nice change from muted and flat pastels.
Error: yek: SHA256 mismatch
Expected: 34896ad65e8ae7c5e93d90e87f15656b67ed5b7596492863d1da80e548ba7301
Actual: 353f4f7467af25b5bceb66bb29d9591ffe8d620d17bf40f6e0e4ec16cd4bd7e7
File: /Users/... Library/Caches/Homebrew/downloads/0308e13c088cb787ece0e33a518cd211773daab9b427649303d79e27bf723e0d--yek-x86_64-apple-darwin.tar.gz
To retry an incomplete download, remove the file above.
Removed & tried again this was the result. Is the SHA256 mismatch a security concern?
Isn't there a Python library that abstracts most of it away with a couple of gigantic classes with incompatible dependencies?
The article doesn't explain how users can exploit these features in UI or prompt. Does anyone have any insight on how to do so?
Very generous to compare to Manhattan Project or CERN.
This explains a lot about the last 1200 years.
We had APL on our Prime minicomputer at Swarthmore when I started there in 1978.
This article is not very good, notably, it is almost all guessing, no evidence.
An explanation that I would have found plausible: building stairs clockwise is cheaper somehow.
The ruling also documents a pattern of Google performing badly in discovery which suggests that it is a strategy.
As the Post article indicates, DOE has tens of thousands of scientists and has worked on NBC weapons for many decades.
DOE supervises the national labs, which have 1) bio experience 2) high classification 3) supercomputers 4) lots more scientists than CIA or God help us FBI. DOE has organic science resources. The other agencies tend to rely on academic consultants, who are compromised because no one wants to bring virology labs under intense and blaming scrutiny.
Switched to Streamlit, haven't looked back.
Hi, I'm putting together an online event called 31 Days of AI for Book-Lovers to coincide with US National Book Month, October 2022. I was struck by the specific call-out to translating literature on your demo page and would like to feature a specifically book-related application of NLLB on one of 'anchor days'. Can someone work with me on this?
Trying to figure out the most expedient way to set it up and test it. Anyone have any ideas?
Microsoft had something called MHTML that did this about 20 years ago ... Tablet PC era.
Key issues with longevity is property rights, land use, and architectural quality. If you build on a parcel that is purchasable,someone will buy it and redevelop. If you choose a spot that is suitable for many uses, one of them will trump residential. If it is ugly or costly to maintain , it will be torn down.
what does this offer that command line tools and status line indicators do not?