HNHacker News
TopNewBestAskShowJobs

_Algernon_

4,630 karma · joined January 1, 2021

submissionscomments
_Algernon_··on Google did not unilaterally decide to kill XSLT
>There's a reason all the major browsers are kind of on-board with this, just like there's a reason it never got updated much beyond XSLT 1. If XSLT would be proposed today, adding it to browsers would be a complete non-starter.

If HTML didn't exist and was proposed today (please forgive the forced hypothetical) it would be a non-starter. Every tech-startup would instead want you to use their dedicated app. We rarely create shared standards or protocols anymore. We should at least keep the ones we have.

_Algernon_··on Being “Confidently Wrong” is holding AI back
Rolling weighted dice repeatedly to generate words isn't factually accurate. More at 11.
_Algernon_··on In a first, Google has released data on how much energy an AI prompt uses
Unless you're arguing that people should kill themselves to save their 100W of power, the future energy use of every alive human is already committed and doesn't matter.

Not to mention that the energy should also include all the extra energy spent on making converting energy into a form that is usable by humans (ie. food). There is probably at least an order of magnitude.

_Algernon_··on Mark Zuckerberg freezes AI hiring amid bubble fears
I'm thinking of exactly what I said[^1] :)

[1]: https://en.wikipedia.org/wiki/Careless_People

_Algernon_··on Mark Zuckerberg freezes AI hiring amid bubble fears
You should read Careless People if this boggles your mind.
_Algernon_··on AI Mode in Search gets new agentic features and expands globally
There are steps to submit it manually: https://github.com/MarginaliaSearch/submit-site-to-marginali...
_Algernon_··on AI Mode in Search gets new agentic features and expands globally
They don't disappear just because Google no longer surfaces them. May I suggest marginalia search for your small web needs? https://marginalia-search.com/
_Algernon_··on Tiny microbe challenges the definition of cellular life
>We can live without bateria if we add with some food supplements.

Citation needed. I would strongly doubt that this is true, because microbes also play a very important role in eg. immune defense. Remove all the mutualitic microbes from a human (eg skin, digestive tract) and the parasitic and pathogenic bacteria will take their place immediately.

_Algernon_··on Vibe Coding Is the Worst Idea of 2025 [video]
>Executives at Nokia and BlackBerry saw the first iPhone, with its lack of a physical keyboard, as an impractical toy for media consumption

This also is how most people use their smart phone today. They were right. Perhaps they just didn't realize that people wanted a on-the-go impractical toy for media consumption.

_Algernon_··on How we exploited CodeRabbit: From simple PR to RCE and write access on 1M repos
Move fast and break things
_Algernon_··on Robots.txt is a suicide note (2011)
I mean the main reason is that robots.txt is pointless these days.

When it was introduced, the web was largely collaborative project within the academic realm. A system based on the honor system worked for the most part.

These days the web is adversarial through and through. A robots.txt file seems like an anachronistic, almost quaint museum piece, reminding us of what once was, while we stoop head first into tech feudalism.

_Algernon_··on Google admits anti-competitive conduct involving Google Search in Australia
Add a multiplier so that doing the activity is discouraged, not just neutral in terms of expected value.
_Algernon_··on The beauty of a text only webpage
The (non-compliant) cookie banner covering half the screen kinda ruins the mood.
_Algernon_··on Blood oxygen monitoring returning to Apple Watch in the US
They didn't poach employees. They made a better offer — in other words competed — in the labor market.

Employees aren't animals in the forest where the king has the sole right to hunt them for sport.

_Algernon_··on Time to End Roundtripping by Big Pharma
Similes don't exist in Europe?
_Algernon_··on What does Palantir actually do?
I thought McKinsey's reason to be was that they sold justification for unpopular decisions management already wants to make.
_Algernon_··on What does Palantir actually do?
Any non-defensive war is a choice. In otherwords every war the US has fought since the war for independence has been a choice (yes, this includes the WWs).
_Algernon_··on Why are there so many rationalist cults?
He writes scare-mongering books about AI doomerism such as If Anyone Builds It, Everyone Dies.

I short, another variant of commercializing the human fear response.

_Algernon_··on Wikipedia loses challenge against Online Safety Act
All your arguments are technical. It's the social layer that is the problem.

>Any well designed privacy system does not rely on the server components doing the right thing.

This is more expensive than just throwing everything into a centralized database. The extra costs needs to be justified when explaining the price to the voters.

>Servers and providers and governments are the main threat actors to be defended against.

Agreed. And they are the ones implementing the system. Clear conflict of interest.

>As for "the purpose can change" well - so? That is also true or anything else, it does not seem like a reason to avoid having good protection now. Any change that could compromise that would not be undetectable - the fundamental crypto should not allow it.

Introducing an age-verification system requires a lot of political capital (as seen by the repeated failures of introducing it so far). Nudging an existing age verification system to cover new purposes requires far less political capital.

>We would know if it happened.

Only if every technical decision goes the right way, despite all incentives and conflicts of interests pointing the other way. I wouldn't bet on it.

_Algernon_··on Wikipedia loses challenge against Online Safety Act
Removing the ability to install a custom OS is just as big of an erosion of rights if not bigger.

The war on general purpose computing is still going strong, if this is suddenly considered the lesser evil.

_Algernon_··on Wikipedia loses challenge against Online Safety Act
What stops the under age user from setting isAdult = 1?
_Algernon_··on Wikipedia loses challenge against Online Safety Act
>It's easy to demonstrate (3) for an age verification system - practical experience will amply demonstrate it to everyone.

No. Absence of evidence that I am not anonymous does not constitute evidence that I am anonymous. Verifiable unlinkability is also difficult to prove.

It may be possible to create a system like this technically, but all social and economic incentives that exist are directed against it:

- An anonymous system is likely more expensive.

- The public generally does not care about privacy, until they are personally affected.

- You have no idea as a user whether the server components do what they say they are doing. Even if audited, it could change tomorrow.

- Once in place its purpose can change. Can you guarantee that the next government will not want to modify this system to make identification of dissenters, protestors or journalists easier?

_Algernon_··on Wikipedia loses challenge against Online Safety Act
Add to that 3) Verifiable to a lay person that the system truly has those properties, with no possibility of suddenly being altered to no longer have those properties without it exceedingly obvious.

This whole concept runs into similar issues as digital voting systems. You don't need to just be anonymous, but it must be verifiably and obviously so — even to a lay person (read your grandma with dementia who has never touched a computer in her life). It must be impossible to make changes to the system that remove these properties without users immediately notice.

The only reason why paper identification has close to anonymous properties is the fallibility of human memory. You won't make a computer with those properties.

_Algernon_··on Wikipedia loses challenge against Online Safety Act
The random article button uses algorithms to decide what content to show to the user.
_Algernon_··on Fight Chat Control
The country that had prosecution of political opponents during the Red Scare and patriot act is better? The US is the country that started all this privacy invading BS and is still way ahead on that front.

After all Chat Control hasn't actually passed yet.

_Algernon_··on Fight Chat Control
Moving to most of those countries would instantly at least 5x (Colombia would be 30x!) my chance of getting murdered (and I assume increase my risk of being the victim of other violent crimes similarly). Not to mention that suggesting El Salvador — a country that has imprisoned 1.7 % of its population, many without being convicted in a court of law — is a truly laughable suggestion.
_Algernon_··on Fight Chat Control
Genuinely curious where you would suggest going. The US isn't better and has been doing this shit since the patriot act.
_Algernon_··on Fight Chat Control
When (rational) people make decisions they weigh the possible rewards of success against the possible costs of failure. We are in a situation where the costs are virtually zero ("oh no, we have to try again in 6 months!") while the rewards are immense: the potential to consolidate even more power to the rich and powerful elite.

It shouldn't be surprising that this happens again and again, and they only need to succeed once. Social movements of the past understood this well. They increased the costs to such an extent that they couldn't be ignored.

Look at the movements that brought forth societal change in the past and imitate them. I can't think of one that didn't have an "extremist" wing that was willing to target the decision makers were it hurt: economic output (eg. strikes or sabotage) and violence.

_Algernon_··on ChatGPT Agent – EU Launch
Yes. This is in essence the same taxonomy used in A Modern Approach.
_Algernon_··on ChatGPT Agent – EU Launch
I'm responding to this claim:

>Agent originally meant an ai that made decisions to optimize some utility function.

That's not what agents originally referred to, and I don't understand how your circling back to LLMs is relevant to the original definition of agent?

← PreviousPage 3 of 34Next →