HNHacker News
TopNewBestAskShowJobs

whacked_new

1,729 karma · joined March 19, 2007

brain & neuroeconomics -> hpc -> semiconductor

en/zh/ja/emacs/vim

mainly python, clojure, typescript

whacked on gh

also whacked on hn but that one's now a placeholder

submissionscomments
whacked_new··on Braille Is Alive, Well, and Ever-Evolving
I'm sighted, but bought a used braille display for slightly less than $200 to learn braille for fun. It's pretty impressive to actuate hundreds of tiny dots at a time, for every sentence you need to read. Every once in a while one dot doesn't drop back when the display is refreshed, and since you're swiping your fingers across the display, feeling (or maybe I was seeing?) that one oddly placed dot is disorienting.

Maybe that feeling goes away with fluency, which I can't speak to, but I have a lot of respect for the engineering that goes into these displays. Granted, I'm not a hardware person.

whacked_new··on Braille Is Alive, Well, and Ever-Evolving
Seems like I went down a similar path. Emacs user, brltty, digital display, the shebang, but not necessarily together :-)

Maybe my experience will be useful to you. Do it for the plasticity.

a high level writeup is here: https://github.com/whacked/cow/blob/main/learning%20braille....

can talk separately in more detail, although for my purposes the code parts are pretty minimal. Brltty does a good job, and so did eBay for me when I started.

whacked_new··on Two Approaches to Decoupling
Most of the discussions about approaches are about shifting the locus of "common language" at a particular point between the producer and the consumer. The common language is basically a codified description of the data that is highly (ideally always) predictable, so that the producer and consumer both speak that language at all times at the boundary of information exchange between the 2 systems.

Something like RPC nudges towards the producer (or backend); graphql moves it to the consumer (or frontend).

A different way is to not look at this as a line of information flow between 2 points, by introducing a 3rd party, an impartial standards body. Maybe it's just a semantic or procedural difference, but when the locus is defined at the meta-architecture level there's no more room for discussion about exactly where you place the locus. Pay the vendor before you ship, or pay the courier when you receive? Neither: introduce an escrow.

whacked_new··on Emergent abilities of large language models
I don't have a good answer, but suspect it will involve a metacognition that can delineate internal and external knowledge.

To further handwave this thought bubble, you can attend to certain thought processes, and intuit the bounds to those processes. An example is if you ask a gifted young chess player, they would say "it felt right", or when asked how Ramanujan figured out a solution to math problems he would say it was divine revelation. In these cases, they know "they did it" but cannot explain how. As the child gets older and learns formal theory, they would give a mixture of logic and inspiration. Both parts can be attended to; the logic will be apparent and the intuition will be revealed, and have different metacognitive properties.

If I were to take a first crack at this, I'd test GPT-AC for its awareness of bounds of its own cognition, not in the sense of "are you able to produce this answer" (because anything any human can come up with, it can answer), but to describe its internal process. Perhaps the language at its intellectual level will be far beyond humans but the language in its metacognitive level will remain primitive.

I don't know though. If our only interface is human language, this is probably a lost cause. But I doubt I will feel guilt in pulling the power cord and maybe that makes me a callous committer of cybercide.

whacked_new··on Emergent abilities of large language models
I have a feeling that based on these emergent abilities, at some point, a sufficiently large multi-modal model will demonstrate a semblance of self consciousness (by way of answering all the designed queries correctly), and the world at large will hail this great achievement, and some hell breaks loose.

And I suspect the semblance of self consciousness is actually a semblance. It will quack like self consciousness but is not actually self conscious.

Just a feeling. But particularly after seeing how these phenomena are organized this way, I can't see how continual emergence of increasingly sophisticated rearrangement and reproduction of information will actually lead to consciousness.

whacked_new··on React is holding me hostage
Having used reagent, watching the React space evolve is like a bricklayer watching a constant stream of innovations around mud. Sure, you could build foundations with bricks, but why do that when you can use a fancy mold and pour mud into it? Why have clean blocks when you can have sticky, messy, dirty goop?

Alas, castles and fortresses have been built with messy goop, and the amount of goop-specific tools and goop slappers far outweigh the bricklayers. So I too slap goop.

whacked_new··on How Duolingo reignited user growth
I have a question to the experienced PMs out there.

If Lenny now time slipped back to the early discussions when they were rolling out the counter and the referral program, would he then vehemently oppose the ideas, citing all that he's learned? "Duolingo isn't like a match-3 game" "Language learners aren't referring discounted rides"

Maybe he would? Maybe he should, because he's learned these lessons. But was younger Lenny's job to be the whippersnapper who says with conviction "you don't know if you don't try, let's give it a shot!"

And so in a sense you must have these two types of people, and once you've been through the gauntlet, you either make it, and have been right all along, or you become wise, and your job is now to be the naysayer.

Or, if you're an experienced PM, are you instead thinking, this isn't perfect hindsight, but simply a largely avoidable detour before he arrived at the metrics that mattered? Maybe a critique of the actual journey is even more educational.

whacked_new··on Some insects I found inside dried Turkish figs from Trader Joe’s
I used to buy and eat organic brussel sprouts from Trader Joe's. Then one day, for some reason, I decided to take a closer look, and found a little green worm crawling about. After removing it, I found another, and another, until I lost both the count and my appetite.

I read that it's normal for brussel sprouts to have these kinds of worms in them, so I must have eaten lots and lots of them. Shrugs, but I can't stand the sight of the worms, so I have never bought brussel sprouts, anywhere, ever since.

I still eat them at restaurants. But only if they're fried or grilled, because that decreases the chances that I'll see a worm.

Unfortunately thanks to this article I'll probably avoid dried figs now because I can't disassociate the visuals.

whacked_new··on Lisping at JPL Revisited
Yea, for the most part.

The problems you describe with external data are supposed to be clojure's forte, but clojure also makes cheap footguns because maps fill most of the use cases for alists / hash tables, and implicit nil punning is everywhere. Once data comes through the wire and is turned into a map, if there isn't an enforceable structure in code, I have to run the program to check the data's shape. Hardly new in dynamic languages, and there are many solutions (e.g. spec, malli, schema, just to name 3 for clojure), but I haven't seen one matching the ergonomics that TypeScript 3+ offers today.

Thanks again! Always enjoy insight from experienced lispers.

whacked_new··on Lisping at JPL Revisited
Thanks for answering my earlier question. Here's hoping you catch another one!

In a lot of my clojure scripts, I forget the shape of the data, and so to help future me, I've use tricks like pre/post constraints, defstructs, schemas, etc. In my experience, TypeScript has the best typing ergonomics (block-local, partial interfaces!). But since I prefer lisp language and interaction ergonomics, I'm always on the lookout for new developments, like https://github.com/coalton-lang/coalton, which has pushed me to start playing with sbcl.

Maybe CCL has already solved this, but I'm not versed in CL, so what methods do you find most effective for passing down knowledge of constraints and data shape to your future selves?

whacked_new··on Lisping at JPL Revisited
When reading your section on Python and ending statements with "pass", and after not finding other references to it in your blog, I have to ask about Ruby. It seems like it would check many boxes (highly extensible, no significant whitespace, begin-end blocks, quite lispy...). If you've played with Ruby, what drawbacks have you found?

What are the major advantages you found in Python and in what cases do they exceed those in CL (aside of e.g. the type system you wrote about in 2008)?

whacked_new··on OpenAI starts testing ChatGPT premium
I think that's a generic response it gives when it's overloaded. I usually just use you.com's search. But every once in a while I'll ask it to make summary tables of stuff. Oftentimes you'd need to fact check the results, but depending on the query you can get some strikingly good answers and clarifications.

It feels like what you get when you slap a REPL onto a search engine.

whacked_new··on OpenAI starts testing ChatGPT premium
I've been using you.com's chat feature a lot recently, which is a ChatGPT-like conversational AI.

It's so good that I have now switched over my default search engine to you.com, as of last week. First change to my default in over 15 years.

whacked_new··on Usbc.wtf – an article and quiz to find the right USB-C cable
thanks for running through the thought experiment, this is pretty clever!
whacked_new··on Liquid democracy: two experiments on delegation in voting
a side note to my sibling post about the bayesian truth serum, which can be used to surface "experts", is that BTS discovers experts by rarity of knowledge in the crowd and correctness of metaknowledge of the crowd. The existence of experts is a critical assumption. A corollary is that the maximal knowledge is bound by the crowd.

But also, since it's relative to the population, the knowledge of "experts" does not need to have any bearing on truth. I thought this would be a problem in our context, but then I thought, that's not so far from reality anyway!

whacked_new··on Liquid democracy: two experiments on delegation in voting
this is indeed a costly problem, but I suspect there are economical solutions using technology.

Expert identification can, to a degree, be bootstrapped out of a crowd using something like Prelec's Bayesian Truth Serum (https://www.science.org/doi/abs/10.1126/science.1102081). The tldr is that it fishes out holders of unusually surprising knowledge relative to the population. I was involved in a study that used BTS effectively to identify expert dermatologists.

I imagine it's possible to find a sweet spot between voluntary querying (a-la slashdot) by a rotating, small party combined with a variant of BTS which surfaces the experts as needed.

The cost of identification will still scale with the group, but it might be doable at ratios that are similar to moderator:users in forums.

whacked_new··on Usbc.wtf – an article and quiz to find the right USB-C cable
The quiz is simple and to the point, with no cruft, so the overall experience is pretty good.

But I have a (slightly tangential) UX/design question: if now you are compelled to remove the Next button from the question pages, how would you achieve it, without reduction of functionality?

For example, on the first question, which is a single-choice question, if you make it submit on-click, then the user would lose the ability to change their mind before submitting, which is reduction of functionality. So maybe you'd consider doing a swipe-to-submit (punishing mouse users).

Then you have multiple-choice questions. Has anyone seen something clever that is both intuitive and removes the need for extra travel to the next button?

whacked_new··on Skribilo: Document Programming Framework
on the surface this looks like a peer of https://docs.racket-lang.org/pollen/

many years ago I would have eagerly given this a whirl, but either I'm too old to try new things, or the broader user ecosystem has convinced me that in order to gain significant mindshare, the syntax cannot stray too far from markdown. Scheme in markdown would be an interesting modification though.

Also, while it tickles the heart of a lisper, one uneasiness I have with systems like these is the lack of easy schema definitions, like the url structure in the skribilo examples.

whacked_new··on Homoiconic Spreadsheets: What, How and Why [video]
Thanks for the reply.

> As to your question, I interpret it as: why are lambdas coming inside spreadsheets instead of spreadsheets becoming incorporated into programming languages, right?

Let me extend this. The power of the spreadsheet programming model (and work environment) is indisputable. In recent years, we've seen many interesting attempts to bring that power directly into programming tools (examples in the Clojure space: witheve, javelin, hyperfiddle; I even think witheve's first iteration was in fact a programmable spreadsheet, but I can't find it anymore). Conceptually these tools have the potential to converge to the same sweet spot, being a merge of 2 similar tenets. But the programming-driven side is converging far more slowly.

I'm interested in this because I think most programming environments these days are still too serial, and there's a lot to gain from moving into a more 2D -- spreadsheet like -- programming environment. It seems obvious, but still far away. It looks like spreadsheets will reach the sweet spot faster. But I like navigating code in Emacs more than I do jumping around cells and hitting F2 to edit.

whacked_new··on Homoiconic Spreadsheets: What, How and Why [video]
True. And they are used by orders of magnitude more people because they have proven their usefulness with the masses

Stating the obvious, but it's a good reminder, because I asked the question anyway. For the mass market, it's a UX problem before it's a product problem before it's a tech problem.

whacked_new··on Homoiconic Spreadsheets: What, How and Why [video]
intriguing presentation! So I have 1 prediction and 1 question.

prediction: even though lambdas bring "code programming" into spreadsheets, and it's a huge leap in functionality for the spreadsheet, the dominant programming mode in the spreadsheet will still be interactive programming in the "traditional" spreadsheet sense. I claim this because the mental model of code programming is different from interactive programming. That said, this will extend the reach of code programming output to the hands of more millions (who will simply copy and paste).

question: the direction of spreadsheets moving into lamdas makes as much sense as "lambdas" into spreadsheets. We've seen several new spreadsheet implementations pop up in the past years, including ones with built-in programming capability, but none of them really seem like as big of a deal as Gsheets/Excel adding lambdas. Presumably this is because spreadsheets include a ton of other features that prove its widespread usefulness even before adding things like metaprogramming. Is this about right, or did I miss a great product/language in the wild?

whacked_new··on Ask HN: Concepts that clicked only years after you first encountered them?
I love this as a research topic. Here are a few examples close to me.

1. The nature of variance in regression methods. The first time I heard about "soaking up the variance" in modifying stat models in brain imaging I had no idea what was going on. Then I spent a few years doing brain imaging and modified models to "soak up the variance" differently. Over the few years I came to grok the concept. I had a electrical engineering PhD colleague with an impressive resume who would argue with me about the effects of models. I realized that he knew the textbook stuff (which I didn't), but he didn't actually grok the concepts (which I did).

2. Once in an office was I mystified by hot spots and cold spots of wifi signal. One of my colleagues, a brilliant engineer, then explained to me what might affect the shape of the hot and cold spots, which is why he sits in certain places. I asked him if he had an intuition of how RF fields are distributed and he said yes, so I asked him how. He said, "I used to do a bit of tensor calculus".

whacked_new··on Ask HN: What was the best software that you used during 2022?
one combination I came to really love this year is babashka (https://github.com/babashka/babashka) + websocat (https://github.com/vi/websocat). I wrote about a method of live web programming with this pair at https://github.com/whacked/cow/blob/main/a%20technique%20for...

babashka isn't strictly necessary; you can also pipe plain text, but pushing hiccup expressions to the browser DOM from the REPL with instant feedback has opened a new world of interactive programming for me.

whacked_new··on Observable Plot 0.6.1
wow, thanks for sharing that post. how many revisions was that? you're a hell of a writer.
whacked_new··on Can Duolingo actually teach you Spanish?
Recently I encountered a tourist who didn't understand anything I said, but said he's from Korea. Recalling that I have been studying Korean on Duolingo for over a year, I excitedly probed my memory for a few seconds, then finally exclaimed, "annyeonghaseyo!" (hello), then blurted in English, "sorry, I don't speak Korean"

Later I encountered a fellow who spoke English. He told me he's from Germany. Recalling that I have been studying German on Duolingo for over a year, I excitedly probed my memory for a few seconds, and finally came up with "sehr gut!" (very good), then "sorry, that's all the German I know"

Then I was at a pizza shop and realized the person at the counter only spoke Spanish. Recalling that I have been studying... finally she lost her patience and thrusted her phone at me, with Google Translate.

True story. I still use Duolingo everyday. But surely, it isn't enough to "know"

whacked_new··on ChatGPT, the Abacus, and Education
I spent a good part of last year picking up the abacus and meticulously tracking my progress. I can attest to "continue to visualize the abacus in their mind, improving their skills even when the physical tool is unavailable", but make no mistake, this is a domain-specific skill which has little bearing on mathematical knowledge and intuition per-se (at least not at beginner-intermediate levels).

One useful aspect which is not often discussed, but might actually complement OP's article, is that the mental abacus converts one kind of mental process (numerical perception) to another (visual scratchpad). Kind of like turning text into a QR-code as a more "efficient" 2D-storage, yet the QR-code contains no "text", and you have to convert it back.

This complementary process in the brain becomes interesting when our mental abilities decline or get replaced, and we might as well re-purpose those brain areas using some kind of transfer encoding (e.g. Morse code).

A naive first-guess could be to learn lists of logical challenges and strategies in order to clarify results given by an AI in the fastest way possible. So given a GAN, train the human on A, and leave the G and N to the computer.

whacked_new··on Babashka is a fast-starting scripting environment for Clojure
babashka is so good it's mind-boggling.

super-easy shell interop + instant nREPL = replaces any complex bash script, driven from emacs!

what's more, you can namespace your scripts and import them. functionalizing code makes scripts easy to refactor, or even trivially move into a full Clojure backend or ClojureScript frontend. With squint-cljs you can even generate self-contained JavaScript directly from babashka code.

whacked_new··on The sad story of Heisenberg's doctoral oral exam (1998)
Many years ago I worked for a professor who allowed me into weekly meetings which were mostly of touring faculty, presenters, or academics on the job market.

The audience was always a pretty stable group of multidisciplinary, world-class faculty. Most of them were cordial but there were a few aggressive ones too. I remember a statistician who drove real hard. I didn't know any of the words he said, but the room felt electrified every time he started.

It was like intellectual UFC, but hero vs mob. I felt for the speakers, but as a fly on the wall, it was amazing. I haven't experienced anything like it since.

whacked_new··on Caterpillar offers phone with built-in FLIR camera
completely fair, although the BV has a layer of anti-slip rubber on the back panel (which I also came to appreciate), so replacing the battery will probably require ripping that layer off (see https://www.youtube.com/watch?v=zikXifsPDT0 if curious)

that said, I've abused the phone enough that maybe I should find a replacement for the rubber too.

whacked_new··on Caterpillar offers phone with built-in FLIR camera
I've been using a BlackView 9900 Pro for over a year, and attest to the surprising usefulness of the built-in thermal camera, such as checking how well food is heated after microwaving, or even watching the heat distribution of rolling, boiled water.

My phone's battery is degraded now and I'm dreading a switch to something else. I've eyed the S62 but don't think it's an upgrade.

← PreviousPage 3 of 33Next →