HNHacker News
TopNewBestAskShowJobs

perrygeo

3,195 karma · joined April 28, 2014

submissionscomments
perrygeo··on A.I. Will Destroy Critical Thinking in K-12
> Even seventh graders can see artificial intelligence is a lesser form of care and attention.

Hard disagree. They quite reasonably view mindless bullshit assignments as bullshit. AI introduced as a tool to produce more bullshit just makes it an extension of a failed model of education. This is what they see. It reflects more on our education system, which seems hellbent on automating and standardizing everything to the detriment of real learning, than on the students themselves and their relationship to AI.

I've watched my son and other middle-schoolers use AI for a variety of school projects. I'll just say that we adults have a seriously skewed misunderstanding about AI's usefulness. Beginner's mind can reveal some foundational truths.

From acting as a muse on essays, to exploring logic/math/philosophy problems, to challenging the LLM foundations (trying to get reasoning models into an infinite loop is a popular game), to scaffolding and debugging pygame code, to developing creative personas, to straight up middle-school level humor. Kids are uniquely able to see its flaws and poke at them, and truly understand how the tech could fit in. No hype, just real lived experiences.

IMO kids see the potential clearly and without the hype that the industry has put on it. They're neither in awe or in fear of the technology, as it should be. The so-called adults in the room should take notice (but they won't).

perrygeo··on Show HN: Hardtime.nvim – break bad habits and master Vim motions
Exactly. h or j repeated is widely dismissed as inefficient. But that ignores two important things:

First, you can hold down h or j and use visual clues to know when to stop. It's one keypress on the homerow, no reason to even glance at the keyboard. Eyes on the monitor = better focus. One keystroke.

Second, thinking in line numbers is one more piece of context to load up into your brain (find the line number), then act on (type out incantation), then context switch back to whatever you were doing before. By contrast, holding j until you're in the spot takes up almost zero working memory. You can think while it's happening! It's a small and subtle difference but that 200ms pause in thought process really does add up - not in wall time (jumping by line is probably "faster") but in contiguous focus time. Having to invoke any multi-part command is a distraction from your core work. Again, one lazy keystroke with immediate visual feedback > some complex combination of keystrokes.

An one small pitch for treesitter - rather than use j/k to navigate by line, set up some keybindings to jump by treesitter element (functions, methods, etc). I've got mine set to +/-. This lets you move in the same "dumb scrolling" way but by jumping to semantically-relevant parts automatically rather than line by line.

perrygeo··on Show HN: Hardtime.nvim – break bad habits and master Vim motions
Funny, I was thinking exactly the opposite. Stack traces, logs, test failures, warning, etc will always reference an absolute number. If you want to tell a coworker where to find a section of code, you use the absolute number.

Relative numbers are useful to save a few keystrokes. But in every other context, absolute numbers are what's relevant.

perrygeo··on “Streaming vs. Batch” Is a Wrong Dichotomy, and I Think It's Confusing
I think I get the analogy, something like: you can append to the record but everything is still record-based? And what do garden gnomes have to do with it :-)
perrygeo··on If nothing is curated, how do we find things
When TV came to American homes in the 1950s, it was a revolution in our national shared consciousness (for better or worse). Obviously there are problems with this - it gives the advertisers and businesses enormous unchecked power to shape society. But we've likely never seen so many people so deeply in sync with the dominant cultural messages.

When streaming became the norm, that dynamic was destroyed. We lurched back to private media consumption (for better or worse). There is no shared cultural narrative to tune into at 8:00 each night. There's millions of disparate voices, screaming into the void 24/7. More freedom and diversity for sure, but nothing coherent you can point to as a culture.

perrygeo··on Show HN: Hardtime.nvim – break bad habits and master Vim motions
Despite using vim and neovim for over 20 years, I'm still apparently a caveman when it comes to navigating. I both love and hate that this plugin calls me out for it. The "Hardtime Report" is a great feature, really shows just how pervasive my bad habits are. This could either help me improve - or force me to uninstall it in anger.
perrygeo··on Thoughts on thinking
Good point. It's not that AI is "pushing us" towards anything. AI can be a muse that elevates our creativity. IF we use it that way. But do we use it that way? I think there will be some who do.

The majority of users seem to want convenience at any expense. Most are unconcerned with a loss of agency, almost enthusiastic about it if it removes the labor of thinking.

perrygeo··on Refactoring Clojure
S-expressions really are the game changer. Especially in a functional lisp with immutable data structures. You can lift and shift forms around with reckless abandon and never cause a bug. You're not typing characters that eventually get parsed into a tree, you're building the tree! Programming in bespoke arrangements of ascii delimiters feels like regressing to punch cards.
perrygeo··on I've never been so conflicted about a technology
I get that. I just don't get why you frame it as an either/or question as though they were mutually exclusive. We can reduce data center emissions, and improve tire maintenance. Por que no los dos?
perrygeo··on Migrating to Postgres
Mid 6 figure DB bill, let's estimate $500k. Divided into 100 million rows (ignore the rest, because db provisioning is typically dominated by the needs of a few core tables). They get 200 rows per dollar.

Your table on a small VPS (which I concur is totally reasonable, am running something similar myself): Let's say your VPS costs $40/mo x 12 = $480/yr. Divide into 150 million. You get 312,500 rows per dollar.

I'd wager you server was faster under normal load too. But is it webscale? /s

There's waste, then there's "3 orders of magnitude" waste. The pain is self-inflicted. Unless you have actual requirements that warrant a complex distributed database, you should "just use postgres".

And just to calibrate everyone's expectations, I've seen a standard prod setup using open source postgres on AWS EC2s (1 primary, 2 replicas, 1 haproxy+pgbouncer box to load balance queries) that cost ~ $700k annually. This system was capable of handling 1.2 million rows inserted per second, while simultaneously serving thousands of read queries/s from hundreds of internal apps across the enterprise. The cost effectiveness in their case came out to ~ 20k rows per dollar, lower than your VPS since the replicas and connection pooling eat into the budget. But still: 2 orders of magnitude more cost effective than the hosted distributed hotness.

perrygeo··on I've never been so conflicted about a technology
I don't understand the "whatboutism" angle. Running LLMs requires massive amounts of electricity and water. Period.

That beef consumption and other activities also require massive amounts of resources is independent. Both can be true. And both need to be addressed. We no longer have the luxury to pick and choose which climate interventions we find most convenient; we need to succeed on all fronts. If the patient is dehydrated and bleeding, do doctors sit around debating if they should give them water or gauze? No, they do both immediately.

perrygeo··on Developers, don't despair, big tech and AI hype is off the rails again
Offshore workers and generative AI have a lot in common. Little formal training. Lots of book smarts but zero context. Performs ok at well specified tasks, extremely poorly otherwise. Cannot understand human-written design documents with enough nuance. No aesthetic design sense. And finally their undisputed ability to pump out high volumes of code (including high volumes of garbage).

The only difference is that LLMs have a deeper and wider understanding of English, no time zone barriers, and nearly instant response time. I find it hard to picture a world where AI doesn't decimate these types of low-skill software jobs.

perrygeo··on The A.I. Radiologist Will Not Be with You Soon
Wonderful insight that I'd never considered. Talk to almost anyone in America and they'll tell you about a health issue that they or their family are deferring due to lack of access. Waiting months or years to simply talk to a specialist, let alone find one that can help, is sadly the norm. Patients rightly feel it's a waste of their time so won't even seek treatment.

Remove that barrier to access and we won't see a shiny new streamlined medical system but rather a flood of new patients requiring even more bureaucracy to manage.

perrygeo··on FDA Moves to Take Fluoride Supplements for Children Off the Market
I'm not sure I see your point. You rewrote that sentence to prove some point about its absurdity, but I find it a perfectly valid statement.

First, it's not a great analogy. Fluoride isn't a psychoactive drug. It doesn't change behavior, it's not addictive, it's not outright toxic to the body at normal levels. It's presence doesn't cause a public health hazard (though its absence might). There is a substantial difference that can't be hand-waved away. We're talking apples and oranges.

Second, have you researched the history of modern nation states and their collective "war on drugs"? I have no qualms saying that it's been an utter failure. Running a centuries long war against its citizens is unconscionable, expensive, and it doesn't even reduce drug use or its impacts.. I see no evidence, none, that militarization of police forces against their citizens has any positive effect on society. It certainty does nothing to stop drug use, which is the entire point.

Put another way, I don't personally drink or smoke (and I happen to think they're far more harmful than we realize), yet I don't want to live in a society where those things are made illegal. There is no contradiction here.

In terms of fluoride, I don't want to live in a society where I'm a) drugged against my will or b) forbidden from accessing the drug entirely. There's got to be a middle ground.

perrygeo··on FDA Moves to Take Fluoride Supplements for Children Off the Market
Astounding. I understand the desire to take Fluoride out of drinking water - involuntarily medicating citizens should be a hard sell for anyone concerned with their legal rights. Fluoride isn't "bad" or "good", it's a exogenous substance and we should have the choice to put it in our bodies or not.

But taking it off the market completely? Come on, that's undermining the entire basis of the anti-fluoride movement. That's effectively saying "fluoride is bad, government says no", which is just a mirror image of the pro-fluoride stance. Just as reactionary and anti-liberty if not more so.

Where are the rational people who want to simply choose (or not) to take a supplement on their own terms?

perrygeo··on How to avoid P hacking
It's not that they suck at statistics. It's that their statistics and experimental designs are artificially stuck in the dark ages. This is forced on the world by the academic publishing industry - you publish this way, or you perish. The completely unsurprising result is a reproducibility crisis that undermines the entire field. Check out "Bernoulli's Fallacy" for a good overview.

My theory isn't that Psychologists are bad at statistics. It's that the remaining problems involve lots of messy interactions and messy data that all but require statistical techniques. We just don't have the tools to extract obvious causality amidst such complexity.

perrygeo··on The great displacement is already well underway?
Opinion: the end of ZIRP has a much greater influence on the job market than AI. No more free money means an entirely different incentive structure. There's a fair bit of "oops we overbuilt in the past assuming we'd have free money to hire more engineers". The interest rate thus mediates how and what we build (Conway's Law strikes again!).

Generative AI is a novelty that makes us crazy productive at certain tasks. But it doesn't yet seem to fundamentally change what we build or why. We just do it faster and sloppier with AI. It's a tactical tool to help you win, whereas interest rates define the rules of the game.

perrygeo··on University of Texas-led team solves a big problem for fusion energy
There are three main hurdles here

First, actually getting fusion to positive energy ROI. That's step zero and we're not even close.

Second, scaling the production of fusion in an safe and economical way. Given the utter economic failure of fission nuclear power (there has never been a profitable one), my priors are that the fusion advocates are vastly underestimating, if not willfully ignoring, this part.

Finally, even if we do get to "too cheap to meter" energy, what then? Limitless electricity is not the same thing as limitless stored energy. Only 20% of our energy needs are supplied by electricity. To wit, the crucial industrial processes required to build the nuclear power plant in the first place can only be accomplished with combustible carbon. A power plant cannot generate the energy to build another power plant. Please let that sink in.

We're already seeing countries with photovoltaic and wind hitting $0/kW on sunny windy days - the grid is nearly saturated for daytime load. There isn't enough demand! This makes the economic feasibility of fusion even less attractive. No one is going to make money from it.

perrygeo··on Continuous glucose monitors reveal variable glucose responses to the same meals
I'm very sensitive to sugar and starch when "at rest". If I wake up first thing in the morning and drink an orange juice or have a bowl of white rice before bed, my blood sugar and mood are out of control. But the same foods mid-day before/after/during intense mental or physical work are very well tolerated if not beneficial. So my priors agree with yours, timing and context absolutely matters.
perrygeo··on Rust’s dependencies are starting to worry me
> the philosophy in Rust is to pull many small package

I'm not sure it's a philosophy, more a pragmatic consideration for compilation speeds. Anyone who's done a non-trivial amount of Rust knows that moment when the project gets too big and needs to split into separate crates. It's kinda sad that you can't organize code according to proper abstractions, many times I feel forced to refactor for compiler performance.

perrygeo··on Throwaway Code: Don't recycle, throw it away (2017)
If you could package this up in a motivational poster, it belongs in every company meeting room. Speed and quality are not two opposing forces to tradeoff. We can have both.

But we need to get rid of this silly, infantile, unwavering attachment to our source code files. Throw code away. All. the. time. The first version of code is, by definition, being built in the absence of critical information. Why on earth would we get so attached to that which was built in ignorance? In this case we're not "reusing code", we're throwing away knowledge!

Why would you discard everything valuable you learned in favor of a code artifact written before you learned it? Throw away the code instead! Surely the code written AFTER gaining the knoweldge will be both faster and better quality. (and more clear, less tech debt, etc)

perrygeo··on CLion Is Now Free for Non-Commercial Use
As good as LSP and other tooling has gotten recently, there are some C++ projects that just need a proprietary IDE for proper code navigation and completion. A lot of times my choice is a) spend 3 hours debugging why neovim's LSP is putting squiggly lines everywhere or b) just fire up CLion.
perrygeo··on White House “Accelerating” Nuclear Power With Executive Orders
> no one wants to live next to a reactor that isn't regulated by the Nuclear Regulatory Commission

Even if given a choice (is that even guaranteed at this point?) some people would chose to fight it, sure. But then you have Starbase, TX that sold their soul to Musk by democratic vote. So there will always be some people who would gladly sell out for anything, including living near a local renegade nuclear reactor, for a chance to make more money.

perrygeo··on Management Habits Burning Out Your Best Engineers
> rewarding them for their lack of skill

I'm not sure it has to do with skill. That would imply the average worker has agency to apply their skills. The dynamic is more: leadership doesn't understand any of technical details and is happy to defer blame on engineers. But engineers are often expressly forbidden from proactively solving problems. So any employee with half a brain can read the writing on the wall - projects will fail due to incompetence (literally no way to stop it) so we might as well prepare to clean up the inevitable mess.

The result: a brittle system that requires heroics just to keep it operating. Once the heroics are rewarded, and proactive problem solving is eliminated, the cycle is locked in place. Cleaning up messes that you yourself begrudgingly create is the new job description.

perrygeo··on I'd rather read the prompt
It's especially damning since cognitive research is not at all ambiguous on the topic. Learning is the result of deliberate practice, intentionally placing yourself in situations where you need to solve problems. The output is irrelevant, building mental capacity is the game. The struggle to learn isn't just some unfortunate obstacle to be optimized away by technology. It is the cognitive mechanism by which we build knowledge.
perrygeo··on AI code is legacy code from day one
Obviously, code is constantly changing. That's not really the point. The point is that as soon as no one understands the code (thus no one on staff to effectively debug or change it) it's "legacy" code.

Let's say you need to make big sweeping changes to a system. There's a big difference if the company has the authors still happily on staff vs. a company that relies on a tangle of code that no one understands (the authors fired 3 layoffs ago). Guess which one has the ability to "shift rapidly"?

perrygeo··on DuckDB is probably the most important geospatial software of the last decade
Mainly a point of clarification - I don't think DuckDB represents anything new in geospatial. It represents a new paradigm for data management and data architecture. Understanding the difference isn't optional.

DuckDB handles floating point numbers too - is DuckDB the most important thing in floating point data? Of course not, the data types and operators haven't changed. The underlying compute environment has. That's where the innovation is. I'd simply appreciate if technical writers took two seconds to make this vital distinction - being precise isn't hard and I don't know what we gain by publishing intentionally muddled technical articles.

perrygeo··on DuckDB is probably the most important geospatial software of the last decade
Why? The article is light on details. Yes, having spatial analysis combined with SQL is awesome and very natural. There's nothing special about 2D geometries that makes them significantly different from floats and strings in an RDBMS perspective - geometry is just another column type, albeit with some special operators and indexes. We've been doing it with PostGIS, Spatialite, etc for two decades at this point.

What DuckDB brings to the table is cloud-native formats. This is standard geospatial functionality attached to an object store instead of a disk. As such, it doesn't require running a database process - data is always "at rest" and available over HTTP. I'm not downplaying the accomplishment, it's really convenient. But know that this is a repackaging of existing tech to work efficiently within a cloud IO environment. If anything, the major innovation of DuckDB is in data management not geospatial per se.

perrygeo··on Ask HN: Anyone else roll eyes at startups that went from "X" to "AI-powered X"?
Has about as much value as "X rewritten in Rust". Which is to say, Rust and AI are both impressive technologies (though very different obviously) and marketing people hope to elevate X by mere word association. Unfortunately, this is an effective scam.

I don't have any problem with labeling yourself an AI business if you're actually objectively working to build AI tech. But if you're vibe coding an app with ChatGPT, you're not building AI-powered technology, you're just consuming it. Might as well talk about "Apple-powered X" if your team is using Macbooks. You don't list the rest of your operating expenses as part of your product, doing so with AI makes you look rather foolish.

perrygeo··on Computational Limit of Life May Be Billion Times Higher Than Assumed
I don't think anyone's disputing that classical neural activity is primary. But the question quickly becomes: what decides and coordinates that neural activity? It is it just an emergent effect of each neuron firing on instinct, producing an output for every input? Or is there some non-local coordination (quantum coherence) that calculates and ensures a correct firing sequence?

Last I checked the evidence was pretty slim for the later. But it's not zero. (Check out Orchestrated Objective Reduction theory). We've seen quantum coherence in microtubules, during photosynthesis, and superradiance in proteins. But no direct evidence of non-local effects in a living brain. Yet.

← PreviousPage 12 of 34Next →