skissane

There's a distinction between theoretical reason–what it is rational to believe, which arguments it is rational to accept or reject; and practical reason–what it is rational to do. Logic is primarily about the former; choosing whether to invest in an IPO is primarily about the later. An action can be theoretically irrational if it is motivated by or involves the acceptance or promotion of irrational beliefs; but making a decision based on "vibes" isn't such a thing.

Have you ever heard the story of Buridan's ass? A perfectly rational donkey placed precisely midway between two bales of hay; it has no logical reason to prefer one to the other, so instead of picking one, it starves to death. That donkey was irrational, not through lack of logic, but through having too much of it. The donkey who picks which bale of hay to eat first based on vibes is more rational than the donkey which completely rejects vibe-based decision making.

> So it seems that you do agree that you're not acting rationally, but you're okay with that as its contained.

No, I don’t agree I’m “not acting rationally”. How do you define “rationality”?

> The article suggests exactly that requiring to delay IPO until 2 consequent quarters of detailed financial audits will be available from Anthropic. Without that there is simply not enough information for an investor to make a rational choice.

I plan to invest a (relatively small) amount of money in Anthropic's IPO. The amount of money I plan to invest is so small (around 1–2% of my total portfolio), the downside is minimal – if it crashes to zero, I've lost 1–2%. I don't see how this is "irrational" for me to do–if I want to do it, the risk is something I'm aware of and can handle. I don't think two quarters of financial results is going to change anything, because I'm doing it based on vibes, not based on stock fundamental analysis.

The norm in a parliamentary system, is the Prime Minister is chosen by the House, and then they (effectively) appoint the other government ministers. If the house doesn’t like their choices, they can always sack the PM. A political party is allowed to adopt internal rules constraining the PM’s freedom of choice in appointing ministers, but the constitution itself doesn’t.

I think for the US, a South African-style system in which the offices of President and Prime Minister are combined, would be more comprehensible to the American public than a (mostly) figurehead President combined with a Prime Minister, as in Germany, India, Ireland, Israel, Malta—the essence of the parliamentary system is subordination of the executive to the legislature, splitting the head of state and government is a complication which adds less value and would be a bigger change for Americans to comprehend.

Australia’s federal constitution says a serving federal minister must be a serving member of the federal Parliament-either the House or the Senate. However, there is a rarely used exception—you can be appointed a minister without sitting in Parliament, but only for a maximum of three months. Part of the intention behind the clause is an Andy Burnham-style scenario in which some outsider is chosen to be PM (or more rarely another minister), and then they can start serving in that role while running in a by-election (special election). This was actually used in 1968 when John Gorton was chosen to succeed Harold Holt as PM, after Holt’s disappearance at sea-Gorton was a Senator, and while it isn’t strictly speaking unconstitutional for a Senator to be PM, by convention they must be a member of the House-so Gorton became PM as a Senator, resigned from the Senate, Gorton then ran for Holt’s House seat and won, becoming a member of the House-all while remaining PM the whole time-so, in the period between his resignation from the Senate and his election to the House, he spent 24 days as PM without being either a member of the House or Senate, but that was under the three month limit. He also spent around 14 days serving as PM from the Senate-which is technically constitutionally allowed, but nobody wants that in practice except as a temporary measure. The original purpose of the three month limit - Australia’s first ministry was sworn in on 1 January 1901, but the first Parliament wasn’t elected until March. (Edmund Barton was chosen to be PM by Lord Hopetoun, whom London had appointed as the first Governor-General; Hopetoun’s first choice was Sir William Lyne, the Premier of NSW; but that choice was very unpopular with both the Australian and British establishments, in response to which Lyne ended up declining the role and recommending Hopetoun appoint Barton instead, which recommendation Hopetoun accepted.)

Australia, Canada and Germany all have federal systems with “snap elections”—a federal system is not an obstacle.

And I think the lack of centralisation of US federal elections is part of what feeds distrust—when every state and even county does something slightly different, when state and local politicians have direct operational responsibility, when you have widespread adoption of voting machines — it is a lot easier for people to believe that fraud is happening. Compare that to the Australian system: federal elections are paper-based, hand-counted, and completely run by an apolitical/non-partisan federal bureaucracy (an independent federal executive agency), which does the same thing nationwide — narratives of distrust in election results find it harder to be accepted, because the system appears more trustworthy.

Adopting a parliamentary system would avoid many of these problems—albeit it would require a constitutional amendment. The President could be elected by the House of Representatives from among its own members, for an indefinite term, but could be sacked by the House at any time by a simple majority vote. The President would continue to be a member of the House, and automatically lose office as President if they lost their seat in the House.

I think part of the issue is-if AI makes you twice as productive at X, that doesn’t necessarily result in any increased revenue, because now your customer expects twice as much for the same price, and if you don’t deliver that one of your competitors will.

A new technology can deliver significant increases in real economic value in terms of utility per $ expended without significantly increasing revenue for the average firm.

> It's been decades since beating the best computer at chess was possible.

LLMs can do an impressive job of writing code that meets some specification-but they still can’t beat humans at deciding what that specification should be. That’s especially true when it comes to non-functional requirements, e.g. what constitutes good coding style or good architectural choices.

> The major difference being that winning a chess game doesn't yield a permanent artifact that is useful, generally.

Chess has a single clearly defined “win condition”; coding doesn’t.

How do you define "harm"? I think there are reasonable definitions by which it would be considered "harm".

Many jurisdictions worldwide criminalise exposing people to pornography without their consent – see for example, Sexual Offences (Scotland) Act 2009 section 6 [0] – and especially doing it to minors. Criminalising doing something to someone without their consent constitutes an official judgement that such an act is harmful.

[0] https://www.legislation.gov.uk/asp/2009/9/section/6

I’m not in the Middle East, but I don’t think the question of the appropriateness of displaying pornography in public places is country-dependent.

> This is hilarious and harmless. Who cares?

Our daughter is 8. As her father, I’d rather she not be exposed to pornography, and I know my wife would agree with me on that. Of course, she’s sure to somehow be introduced to it eventually, but I hope not until she’s significantly older.

I think this is something about the incentive structure of the US legal system.

Here in Australia, a member of my family was involved in a fatal accident with a pedestrian, around 20 years ago. They put him on trial even though the case was weak, and then the prosecutor moved for dismissal part way through the trial. But, what people told us, is prosecutors are encouraged to take weak cases to trial because dropping the case part way through the trial is more politically palatable than refusing to bring it to court to begin with.

The big problem with these permission dialogs, is their lack of technical detail about what the app is actually trying to do which triggers them.

How do I know whether to say "Allow" or "Disallow" when it gives me no idea what either option actually means in concrete?

Well, if someone found a quasipolynomial time algorithm – O(n^log(n)) – for an NP-complete problem, that would technically be compatible with P!=NP, but might have quite significant practical implications.

I think a proof of P=NP via a galactic algorithm is far more likely than a proof of P=NP via an efficient algorithm.

I mean if someone develops an O(n^100) algorithm for solving an NP-complete problem, that would prove P=NP, but pose no immediate threat to cryptography.

It increases risk, in that a galactic algorithm increases the odds that an efficient algorithm exists waiting to be found, but it is entirely (epistemically) possible that the galactic algorithm exists but the efficient algorithm doesn't.

> So the term "guy" became genericized, beginning in Australia it seems, to refer to a grubby lower-class Catholic boy.

I'm sceptical of this alleged Australian connection. As far as I'm aware, the term started in England, spread to the US, and it is in the US that its usage really took off. In Australia, "guy"/"guys" as a term for a male or gender-neutral person was imported from the US in the second half of the 20th century.

Well, it turns out they actually are bothering – while in DDR thru DDR4, the mode registers are write-only, LPDDR2 onwards have an MRR (Mode Register Read) command, which the CPU/chipset can use to inspect the manufacturer and capacity; and DDR5 has copied the same facility from LPDDR.

> RAM sticks have a little tag chip that tells a PC "how much RAM this stick has", but RAM chips themselves don't - so the device has to infer its RAM capacity from elsewhere

I'm surprised that in 2026 this is still true.

You could easily fit a little microcontroller on the die of each RAM chip which reports its make/model/manufacturing date, capacity and capabilities, can be triggered to run self-tests, etc. If they aren't doing that yet, I'm surprised that they aren't.

Well, one option Quad9 could consider, is moving to Liechtenstein - it is part of the EEA (unlike Switzerland which isn’t, although its bilateral agreements with the EU/EFTA partially approximate EEA membership), but unlike Switzerland, it isn’t party to the Lugano Convention

However, I don’t know whether the other aspects of Liechtenstein’s legal and political system would make that viable or not.

As some other comments on this post have mentioned, there is a a treaty known as the "Lugano Convention", whose parties are the EU (as a whole), Denmark [0], Iceland, Norway and Switzerland; post-Brexit, the UK wanted to join in its own right, but the EU refused. Under the Lugano Convention, if you get a judgement in a French court, you can then apply under the convention to have a Swiss court enforce it; however, there are exceptions which the Swiss court could rely on to refuse to do so – but I have no idea what are the odds that a Swiss court would decide any of those exceptions apply.

[0] Denmark has an opt-out from this area of EU law, which is why it is the only EU member state which needs to be directly party to the treaty; it applies to the others indirectly through the EU as a whole being a party to it

> They're trained on all of human history,

Not really… their training datasets strongly skew towards the contemporary West; LLM training data isn’t representative of the contemporary global average, let alone the historical global average.

As well as the skew of the raw training data, RLHF further skews it—a lot of the beliefs of the average 2026 human being (who belongs to the global South) are likely unsafe/misinformation/etc when judged by the standards of the AI vendors.

The MOS 6509 has bank switching hardware built into the CPU, instead of being on an external MMU chip such as the MOS 8722.

What makes the CBM-II's 6509 fundamentally different from the C128's 8502+8722 combo? And what makes either fundamentally different from the segmentation scheme of the 8086/8088?

> The 6502 can't actually address more than that.

Unless you use bank switching, which historically a lot of 6502 machines did use (e.g. Apple III/IIe/IIc, BBC Master 128, Commodore 128, Commodore CBM-II, etc)

> Developers work with certificates a bit too seldom to aquire proficiency with conversions. Being able to use PEM directly will ease a lot of pain.

While I certainly welcome this in the JDK, in practice it hasn't been a huge issue since a lot of people just use Bouncycastle to get that support.

Here is my input: https://gist.github.com/skissane/dfb7a569968224ec98cda757371...

To be clear, the issue is not that it says that text is AI-generated – that text is indeed AI-generated. The issue is the detection reason it gives:

Repeated sentence openers · 3 sentences opening “-”

A nested markdown list is not a reliable sign of AI-generation, humans were using them long before any AI was.

Similarly, its second "AI tell" is:

Echoing sentence runs · 2 sentences echoing “google sso auto create”

Two sentences beginning with the `GOOGLE_SSO_AUTO_CREATE_` is not a sign of AI generation, it is a sign of quoting code

Only the third match is a genuine "sign of AI generation"

In the Australian system, your employer has to pay 12% of your pay into a pension ("superannuation") fund of your choice. Taxpayers still pay for pensions for those whose individual pension accounts have run out – but, in coming decades, that taxpayer expenditure is projected to shrink. [0]

Maybe the US should look at copying it? Well, Trump has expressed interest in doing so. [1] But, Australia has had something like this since the 1990s, so even if the US adopted the Australian model tomorrow, it would take a few decades for it to really make an impact.

[0] https://www.thesenior.com.au/story/9346972/age-pension-spend...

[1] https://www.ncpers.org/blog/what-trumps-interest-in-australi...

I think there is a quantifiable measure-create a test dataset containing questions and human expert-written answers; give the questions to an AI to get corresponding AI-generated answers; then ask humans to guess which answer was human-written and which was AI-generated; now you can measure how much better than chance the humans perform at AI output detection, which gives you a quantifiable measure of the human-likeness of the AI responses.

Suggestions:

- add to your README links to download the relevant .gguf files – I assume from HuggingFace?

- or even implement your own equivalent of `ollama pull`/`list`/`rm`/etc

- add some CI using Github Actions (or an alternative if you prefer)

- as well as `go install github.com/computerex/dlgo/cmd/dlgo@latest`, also document `go install github.com/computerex/dlgo/cmd/dlgo-vision@latest`, `go install github.com/computerex/dlgo/cmd/dlgo-image@latest`, etc (obvious to anyone experienced with Go, not to many inexperienced with it)

actually, I try running `go install github.com/computerex/dlgo/cmd/...@latest` on my Mac and I get a lot of errors; some of them are because `cmd/gpu-watchdog` is trying to use Windows-specific APIs (e.g. `syscall.NewLazyDLL`) even on macOS

Never seen that tool before, but it has an obvious bug: paste in markdown-formatted text with `-` bullet points, and it told me a sign of AI writing was: `Repeated sentence openers · 3 sentences opening “-”`

Now, the text I pasted in so happened to really be AI-generated, but I don't think using markdown bulleted lists is a reliable sign of AI-generation.