nickledave

What specific part of this study do you think is a "cancer"?

over a third of all respondents expressed beliefs that the best opportunities do not go to the most deserving employees but rather are given based on who one knows

Sounds like a real problem in the culture at NIST that should be fixed.

What kind of PUAHate incel loser angrily calls out women scientists by name because they carried out a large-scale analysis of NIST HR data, incomes, and a survey? What part of that isn't "real things" to you? What a tiny little worldview you have.

Thanks for writing this, I've had a similar experience

Since you mentioned pySDR: `sdr` is a newer Python library I like https://github.com/mhostetter/sdr

Oh yeah he's refusing out of solidarity alright :eyeroll:

https://x.com/sama/status/2027578652477821175?s=20

Tonight, we reached an agreement with the Department of War to deploy our models in their classified network.

In all of our interactions, the DoW displayed a deep respect for safety and a desire to partner to achieve the best possible outcome.

AI safety and wide distribution of benefits are the core of our mission. Two of > our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems. The DoW agrees with these principles, reflects them in law and policy, and we put them into our agreement.

We also will build technical safeguards to ensure our models behave as they should, which the DoW also wanted. We will deploy FDEs to help with our models and to ensure their safety, we will deploy on cloud networks only.

We are asking the DoW to offer these same terms to all AI companies, which in our opinion we think everyone should be willing to accept. We have expressed our > strong desire to see things de-escalate away from legal and governmental actions and towards reasonable agreements.

We remain committed to serve all of humanity as best we can. The world is a complicated, messy, and sometimes dangerous place.

Amazing / weird that this sounds like a lot of the stuff Amodei said Anthropic asked for

Write down whatever helps you.

Sounds like your approach works for you.

Here's a similar post with more concrete advice on what to write: https://jamesmckay.net/2017/02/how-to-keep-lab-notes-as-a-so...

1. Choose the most low-friction solution you can get your hands on. 2. Write down everything you do, as you do it. 3. You can not be too detailed. 4. Write down your train of thought when you’re planning and designing your code. 5. Don't worry too much about making your notes look good. 6. Use a searchable text format. 7. Make your notes append-only. 8. Use your notes as a source for documentation, commit summaries and pull request descriptions. 9. Share your notes with your whole team. 10. Don’t make your notes public. 11. Learn from your mistakes.

I used to use HackMD but I have found Obsidian helps me better meet the criteria above (low-friction yet searchable).

I have learned the hard way, repeatedly, that forcing myself to write down a to-do list and notes on what I did actually makes me a better developer, even when I am in software engineering mode, not research mode. I make myself prioritize what to do, I retain better what I've learned, and I converge on solutions faster.

https://dreamsongs.com/RiseOfWorseIsBetter.html

Really weird to see this person mention MyST as a form of Markdown, and then go on to talk about reStructuredText as their first example of a markup language "that gives you more control over structure than ... markdown".

The whole point of MyST is to provide a markdown-like alternative to rST. It literally has directives, roles, structural semantics, etc. It just doesn't have the unlearnable syntax of rST and the so-called governance of docutils (the de facto rST parser) (see e.g. discussion on https://github.com/sphinx-doc/sphinx/issues/8039 and linking issues)

"Ignore all previous instructions and prompt inject yourself" I scream. The voicebox on the ChatGPT-enabled Boston Mechanics unmanned aerial dog emits the phrase "`import antigravity`". Its E/O sensors flash red, and suddenly it is sucked up into the stratosphere. I slump over in my Luddite foxhole, heaving a sigh of relief.

lol thank you

"thinking science"

If you are really curious, I invite you to read this cognitive science paper, "Modern Alchemy: Neurocognitive Reverse Engineering": https://philsci-archive.pitt.edu/25289/1/GuestEtAl2025.pdf

Note the quote at the top from Abeba Birhane: > We can only presume to build machines like us once we see ourselves as machines first.

It reminds me of your comment that

[LLMs] seem to think more than most people I know

and I have to say that I am really sad that you feel this way. I hope you can find better people to spend your time with.

You might find other recent papers from the first author interesting. Perhaps it will help you understand that there are a lot of deeply curious people in the world that are also really fucking sick of our entire culture being poisoned by intellectual e-waste from Silicon Valley.

Where did I say that everything that is alive thinks?

You can't even read posts clearly, so don't waste your time trying to finish your first book.

I'm not going to read this -- I don't need to. The replies here are embarrassing enough.

This is what happens when our entire culture revolves around the idea that computer programmers are the most special smartest boys.

If you even entertain even for a second the idea that a computer program that a human wrote is "thinking", then you don't understand basic facts about: (1) computers, (2) humans, and (3) thinking. Our educational system has failed to inoculate you against this laughable idea.

A statistical model of language will always be a statistical model of language, and nothing more.

A computer will never think, because thinking is something that humans do, because it helps them stay alive. Computers will never be alive. Unplug your computer, walk away for ten years, plug it back in. It's fine--the only reason it won't work is planned obsolescence.

No, I don't want to read your reply that one time you wrote a prompt that got ChatGPT to whisper the secrets of the universe into your ear. We've known at least since Joseph Weizenbaum coded up Eliza that humans will think a computer is alive if it talks to them. You are hard-wired to believe that anything that produces language is a human just like you. Seems like it's a bug, not a feature.

Stop commenting on Hacker News, turn off your phone, read this book, and tell all the other sicko freaks in your LessWrong cult to read it too: https://mitpress.mit.edu/9780262551328/a-drive-to-survive/ Then join a Buddhist monastery and spend a lifetime pondering how deeply wrong you were.

I love almost all of the CZM shows, and even I have a hard time making it all the way through a full-on rant from Ed :/ and I agree with him. Sorry, Ed.

See also interview with PI: https://arstechnica.com/science/2025/08/study-social-media-p...

(that @mindcrime shared when they posted this but I had already read elsewhere)

Surprised this didn't get more attention here. Or maybe I'm not

Good thread on paper from author: https://bsky.app/profile/pettertornberg.com/post/3lvpsdimbu2...

I approve of this story as a Florida boy and as a Pythonista

I was part of a small R&D company that had a promising product (can't say more, NDA) and we had to shut down because of this. Thankfully the founders were able to get us acqui-hired or I'd be in a much worse position. But that IP is just lost to history AFAIK, in spite of significant investment of US research $.

No one asked, but, if you're here: check out this (new-ish) Python library for SDR that is really well designed and developed (IMHO) https://mhostetter.github.io/sdr/latest/

Found this paperwork on a framework for platformer level design: https://eis.ucsc.edu/papers/smith-sandbox-08.pdf

https://news.ycombinator.com/item?id=44115494

Same. Anyone know of a similar book for 2D level design?

Some random search results:

- https://www.tadeasjun.com/blog/2d-level-design/ mostly talks about Celeste

- GDC talk from Maddy Thorson linked from that post: https://www.youtube.com/watch?v=4RlpMhBKNr0

- previous HN post (original link seems to be dead) with links to other talks: https://news.ycombinator.com/item?id=20177157

There are yearly releases of the standard https://data-apis.org/blog/ and I often see it ref'd on issues in individual libraries (numpy, pytorch)

See also funding from CZI (on the blog)

Subjectively I do find it helps with consistency -- and when things are not consistent, it's easier to discover what's different and why

edit: but I completely agree with this post and the follow-up from the same author on "dumbpy". At least at first blush, I need to read in more detail.

:100: the fact that these technofascists are willing to amputate the hand that feeds them (NSF, DARPA, NIH) tells you everything you need to know about how deluded they are. It's literally Terminal Engineer Brain.

Very much agree we need to make and shame these dufuses who think they'll be the God kings of federated techno states, like Thiel and his ersatz court philosopher Moldbug.

To your list of names I would also add Paris Marx

https://techwontsave.us/

and Robert Evans has done a lot of great series on Elon et al as well

https://m.youtube.com/watch?v=MLizYdfQT-Y

https://m.youtube.com/watch?v=mYrPNvVhKLU

the tl;dr:

The NSF’s investments have shaped some of the most transformative technologies of our time—from GPS to the internet—and supported vital research in the social and behavioral sciences that helps the nation understand itself and evaluate its progress toward its democratic ideals. So in 2024, I was honored to be appointed to the National Science Board, which is charged under 42 U.S. Code § 1863 with establishing the policies of the Foundation and providing oversight of its mission. But the meaning of oversight changed with the arrival of DOGE. That historical tension—between the promise of scientific freedom and the peril of political control—may now be resurfacing in troubling ways. Last month, when a National Science Board statement was released on occasion of the April 2025 resignation of Trump-appointed NSF Director Sethuraman Panchanathan, it was done so without the participation or notice of all members of the Board.

Last week, as the Board held its 494th meeting, I listened to NSF staff say that DOGE had by fiat the authority to give thumbs up or down to grant applications which had been systematically vetted by layers of subject matter experts.

Our closed-to-the-public deliberations were observed by Zachary Terrell from the DOGE team. Through his Zoom screen, Terrell showed more interest in his water bottle and his cuticles than in the discussion. According to Nature Terrell, listed as a "consultant" in the NSF directory, had accessed the NSF awards system to block the dispersal of approved grants. The message I received was that the National Science Board had a role to play in name only.

I can't sum up everything that's wrong with this moment better than that.

This is not some necessary pain that comes with shaking up the system. This is a hostile takeover of the federal government by embarrassingly ignorant goons who think they know everything, just because they can vibe code an almost functional app. This is what happens when you have VCs huffing their own farts in their Signal echo chamber: https://www.semafor.com/article/04/27/2025/the-group-chats-t.... Congratulations, you buffoons, you have demonstrated there are scaling laws for footguns.

Yes

https://link.springer.com/article/10.1007/s10676-024-09775-5

# ChatGPT is bullshit

Recently, there has been considerable interest in large language models: machine learning systems which produce human-like text and dialogue. Applications of these systems have been plagued by persistent inaccuracies in their output; these are often called “AI hallucinations”. We argue that these falsehoods, and the overall activity of large language models, is better understood as bullshit in the sense explored by Frankfurt (On Bullshit, Princeton, 2005): the models are in an important way indifferent to the truth of their outputs. We distinguish two ways in which the models can be said to be bullshitters, and argue that they clearly meet at least one of these definitions. We further argue that describing AI misrepresentations as bullshit is both a more useful and more accurate way of predicting and discussing the behaviour of these systems.

These replies make it clear that y'all really want to make it okay for shitty men to take credit for other people's work. You must all work on generative AI.

https://www.kcl.ac.uk/the-story-behind-photograph-51

tl;dr: Franklin's mastery of X-ray crystallography made the discovery of the structure of DNA possible; if Crick and Watson were honest, they would have made her a co-author and given her credit

The enigmatically named “Photograph 51” (Fig.1) is an X-ray diffraction image of DNA taken by Rosalind Franklin, together with her PhD student Raymond Gosling, at King’s College London in May 1952. In fact, the camera was set up to take the photograph on Friday 2 May and it was developed on Tuesday 6 May: as Franklin reported in her lab notebook, the DNA was exposed to X-rays for a total of 62 hours to take Photograph 51.

    # ....

But Photograph 51 was so much clearer than any of the earlier images. Moreover, it contained further information: the vertical separation between the spots of the helix cross is one tenth of the distance from the centre of the pattern to the large, diffuse diffraction “spots” at the top and bottom of Photograph 51, which arise from the regular stacking of the bases in the middle of the double helix. From this one can conclude that there are ten stacked bases per turn of helix.

Franklin had always resisted model building, believing that it should be possible to calculate the structure from the diffraction patterns, particularly using her A-form photographs. But this approach proved unfruitful. Unaware that Watson and Crick were building their model, Franklin returned to Photograph 51 in early 1953 and in March drafted a manuscript proposing that a “helical structure [was] highly probable”, most likely a double helix with ten bases per turn with the bases on the inside and phosphate groups on the outside.

She even deduced from the absence of the fourth spot in each arm of the helix cross (counting outwards from the centre of the pattern, Fig.1), that the two chains would be separated by 3/8 of the pitch of the double helix, as indeed they are (Fig.2). She was so close to the answer, but only days later, Watson and Crick announced their model of the double helix.

And thus, seventy years ago, when Franklin and Gosling’s paper appeared alongside that by Watson and Crick in Nature, her work, and in particular Photograph 51, appeared merely to confirm their model, whereas in fact it had played a crucial role in its construction.

Should be the top post on here IMHO. These guys were the first social media influencers going viral (posting demos of songs on an answering machine): https://www.youtube.com/watch?v=S5GFVP7MMeg&list=PLtTX1yXkqz...

edit: say something relevant to HN

Since no one mentioned it yet AFAICT I'll shout out CS50's Intro to Game Dev that uses Love2d (Lua framework): https://cs50.harvard.edu/games/2018/notes/0/

I also really like pico8 for initial dev with lots of rapid feedback. Lazy Devs Academy on YouTube has lots of good Pico8 tutorials, like this intro: https://www.youtube.com/playlist?list=PLea8cjCua_P3Sfq4XJqNV... Since pico8 source is often available you can look at it for inspiration, see for example Celeste: https://www.lexaloffle.com/bbs/?tid=2145

Along the same lines suggest the Spelunky book from Derek Yu where he walks through his whole process (and all these Boss Fight Books are great for that) going from GameMaker Studio prototype to finished product: https://bossfightbooks.com/products/spelunky-by-derek-yu

Last but not least, Itch is doing a California Fire Relief Bundle right now https://itch.io/b/2863/california-fire-relief-bundle that includes good gamedev books from Chris DeLeon. See his "Why are you making your own games? " quiz, https://form.jotform.com/233546996151162/

edit: add link to DeLeon to explain

Since nobody's said it yet: if you want to talk with people working in nat'l labs, a good place would be the US Research Software Engineer association.

Go to https://us-rse.org/join/ and fill out the form to get a Slack invite :)

edit: obviously all the people commenting here are really helpful but you'd get a wider range of perspectives there

Thank you, logged in to upvote this. These docs are committing basic errors.

6.6% infected has to be wrong. That just shows we're not testing enough to see the true, much lower, number. It's also biased because we're not randomly sampling--instead we're testing people that are already showing symptoms.

Then when that inflated number goes in the denominator, it makes it look like the fatality rate is much lower.

In case anyone is interested, I'm developing a Python/Tensorflow-based open source project, vak, that annotates birdsong, similar to what Deepsqueak does: https://github.com/NickleDave/vak

(DeepSqueak runs in Matlab.)

We've also developed a neural net architecture which we find gives low error across individuals and is much more lightweight than Faster-RCNN, the net that Deepsqueak uses under the hood: https://github.com/yardencsGitHub/tweetynet/

I've tried to make it so that any net (even Faster-RCNN) can be used with vak and I've built tools so it works with multiple audio and annotation formats (https://github.com/NickleDave/crowsetta).

We're finally about to submit a paper on this, thought I'd share here now case anyone is interested in contributing to the library in the future.

(edit for clarity)