0hijinks

I don't have an axe to grind, but this story piqued my interest. So here's some more context:

- SunJun8's original issue text is still available at the linked issue. The gist is that Nous Research's Hermes appears derivative of (or heavily inspired by) EvoMap's Evolver (https://github.com/EvoMap/evolver), but without attribution.

- Issues #17688 and #27266 repeat this request for attribution.

- teknium1 retitled the original issue to "." and edited the issue text to "."

- Nous Research deleted comments from 4 users, including the issue submitter, and blocked all of them.

- No formal response has been given by teknium or the Nous Research project. It appears they are trying their darndest to brush it under the rug.

- EvoMap posted a whole manifesto (https://evomap.ai/blog/hermes-agent-evolver-similarity-analy...) about the similarities.

Plagiarism is bad, even in the age of AI, whether perpetrated by SV or China. If credit is due, give it. Else, all it costs you is your credibility. If credit is not due, make an official statement as such. It's absurd to take (seemingly valid) criticism and sloppily paint over it.

I guess I had never considered at face value what that kind of arrangement means until reading this DOJ document.

It makes total sense to me that committing terrible crimes should be punishable. When a criminal per domestic law resides in a different country (and the issue crosses national borders), it seems to me (not a lawyer) that the determination of appropriate force must be decided by an international group.

Otherwise, what's chilling to me is... what stops a country from unilaterally abducting anyone they want? The country doing the abducting determines the laws that have been violated, after all. In this document, we've explicitly determined that international law has no bearing on domestic consequences. So do whatever you want, so to speak. Freaky stuff.

His rhetoric towards Maduro has a good consistency, yes.

Mind you, his messaging on corruption in government, domestic terrorism, Ukraine, and the free-market economy in the U.S. is all over the place.

It's surreal to read written arguments defending an "extraordinary rendition" using the military in terms of domestic criminal cases. And to so plainly state that international law ought not apply to the leader of the free world, effectively because we think we can and our interests come first, is distressing. If this is an "arrest" of a "fugitive", to keep with the DOJ analogy, it should be subject to the laws of its jurisdiction. Not the laws of whomever has air superiority.

... international law does not prevent [us] ... from ... arrest[ing] individuals [residing in a foreign state] for violations of United States law.

God, that's chilling.

Congress has declined to amend relevant statutes to deny the Executive the ability to engage in rendition.

Congress's continued appropriation of funds to the agencies known to engage in the practice should be taken as (at minimum) acquiescence.

So, because our Congress can't pass a bill handed from God himself unless it fills their coffers, we're silently consenting to this baloney?

And I get that legal precedence is a thing, but we shouldn't be looking at military operations in Haiti, or Libya, or Iraq as justification for more international shit-stirring. It's a slippery slope.

Albeit semantics, rebranding to the War Department is not good optics if one intends receive said department's advice before carrying out "a use of force that ... does not rise to the level of war in a constitutional sense". This nonsense is maddening.

its carcinogenic effects are now well known (no more using it to decaffeinate coffee, degrease engines, or in shaving cream, thankfully!)

I had no idea benzene was used to decaffeinate coffee [1]. Solvent separation makes sense, but I'd never really thought about it. Definitely not a residue I'd want to drink.

From an amateur chemistry perspective, benzene is so fascinating. Benzene rings are one of the first, maybe the very first, structures students learn that don't have fixed, discrete bond counts. That they can be interpreted as a cloud of 1.5-electron bonds between each of the 6 carbons, or alternating double- and single-bonds, makes for some interesting hand analysis of molecules.

[1] https://en.wikipedia.org/wiki/Decaffeination#Direct_method

So you must always factor in the amount of time before a predicted crash: the longer it takes the more money you lose.

I think the idea is, as a put buyer (market taker), this has already been baked into the option premium. The only "maintenance cost" in the sense of a cost that adds to an open position is from interest on margin loans.

There would be a maintenance cost from rolling the position into a later expiry, but I think the impression is that this is a precise single bet.

EDIT: You're spot on about opening a position being a sort of cost too, due to missing out on risk-free returns. This is especially important for hedging. Less so for a directional bet.

> ...we can see that, in less than two decades, Earth has tilted 31.5 inches as a result of pumping groundwater. This equates to .24 inches of sea level rise.

For those confused how they managed that geometric analysis, the sea level rise mentioned in the paper [1] is caused by groundwater depletion. The tilt is caused by groundwater depletion. The sea level rise is not caused by the tilt.

[1] https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL10...

The outage page has been updated to include context:

> MathWorks experienced a ransomware attack. We have notified federal law enforcement of this matter. The attack affected our IT systems. Some of our online applications used by customers became unavailable, and certain internal systems used by staff became unavailable, beginning on Sunday, May 18. We have brought many of these systems back online and are continuing to bring other systems back online with the assistance of cybersecurity experts.

Tinnitus is sometimes neurological, seemingly caused by the brain compensating for a loss of sensation. I can imagine a horror story in which this just makes it a thousand times worse, on top of permanently losing all hearing.

Now, being able to use a hot-swappable audio sensor instead of an ear made of tissue would be pretty dope.

As a user, I'm under the impression YouTube uses click follow-through for algorithm feedback. For the past two years, I've consistently gotten more random content with ~100 views suggested to me in the side bar. I often click and check it out. Maybe prime the pump by diving for some random vids?

From the BOM:

MCU is an STM32H743IGT, external SDRAM is a Winbond W9825G6KH-6, and audio CODEC is a TI TLV320AIC3104IRHBR.

> Unsurprisingly, the combined solver performed the best, solving the puzzle in an average of 4.77 moves. The quantum solver was next, with an average of 5.32 moves, while the classical solver came in last place with 5.88 moves on average.

This effect is pretty neat. From the paper [1], the quantum solver can only do what they call "square root SWAPs", which is like a tile swap that relies on certain superposition rules. The classical solver can only use standard tile SWAPs. The combined solver can do both. A little over half the puzzle states are solved faster by the classical solver, but certain tricky states benefit from this new "move type". So the game had this quantum-like computation option tacked on, but certain initial positions just don't benefit from it.

I don't see any sort of "applications" section in the paper. They talk about how I guess you could build the puzzle thing with "arrays of ultracold atoms in optical lattices", but that still doesn't answer the question. My takeaway is that even problems which benefit dramatically from quantum algorithms in some cases (in a future where that's cheap and widely available) should have careful algorithm design built on other heuristics.

[1] https://arxiv.org/pdf/2410.22287

Links to the article and the PDF both are behind this human test. Guess today is the day I learned I'm a robot.

As all eight Ivy Leagues collectively received this amount in funding in 2024, it would appear that Yale is planning for a longer-term drought.

Aside, I tried to copy the title to search for context about this elsewhere:

> To be able to copy & paste content to share with others please contact us at subscriptions@pei.group to upgrade your subscription to the appropriate licence

Well, that's... something. No ma'am, I don't think I will.

Small M-dwarf stars ... operate through convection ... likened to what we see in a boiling cauldron of water.

Larger stars like the Sun show a mix of radiative transfer – photons being absorbed and reabsorbed as they make their way to the surface – and convection.

That enhances M-dwarf flare activity as their plasma is twisted and rotated, producing magnetic fields that snap open only to reconnect.

This is the first post I've seen targeted toward generic geeks that explained it that way. It makes total sense, is really cool, and I'm glad they wrote this article.

The point about entry into massive, maybe untractable datasets makes sense to me. I get the usefulness of GenAI there for sure. Another commenter suggested:

> Its more like, here are some possible answers where there were none before.

That's a good point. AI doesn't need to be an authority, but another way of generating leads, maybe when it would be time-intensive to do so yourself.

If it's not prohibitive to do the digging with a human, do that. Because if adequately trained and rested, the human will perform more reliably.

It sure seems like the use of GenAI in these scenarios is a detriment rather than a useful tool if, in the end, the operator must interrogate it to a fine enough level of detail that she is satisfied. In the author's Scenario 1:

You upload a protest photo into a tool like Gemini and ask, “Where was this taken?”

It spits out a convincing response: “Paris, near Place de la République.” ...

But a trained eye would notice the signage is Belgian. The license plates are off.

The architecture doesn’t match. You trusted the AI and missed the location by a country.

Okay. So let's say we proceed with the recommendation in the article and interrogate the GenAI tool. "You said the photo was taken in Paris near Place de la République. What clues did you use to decide this?" Say the AI replies, "The signage in the photo appears to be in French. The license plates are of European origin, and the surrounding architecture matches images captured around Place de la République."

How do I know any better? Well, I should probably crosscheck the signage with translation tools. Ah, it's French but some words are Dutch. Okay, so it could be somewhere else in Paris. Let's look into the license plate patterns...

At what point is it just better to do the whole thing yourself? Happy to be proven wrong here, but this same issue comes up time and time again with GenAI involved in discovery/research tasks.

EDIT: Maybe walk through the manual crosschecks hand-in-hand? "I see some of the signage is in Dutch, such as the road marking in the center left of the image. Are you sure this image is near Place de la République?" I have yet to see this play out in an interactive session. Maybe there's a recorded one out there...

That's kind of a bummer, but still neat to have built-in self-hosting. I think I've seen videos of perfboard computers that allow manual data entry with pushbuttons and manual clocks. I could see that being extended to punch cards. (Not trying to be flippant; I think that could be an interesting extension to this sort of project. Have a bucket of parts, a perfboard, and some paper? Let's flash an operating system.)

This. Skill sharing among a small tribe is the most effective way to survive a catastrophe, and I am concerned modern societies are poorly conditioned to do it well.

Speaking from a North American perspective, kids are educated in how to succeed in a national/global economy, not how to build small communities and develop/share useful skills. TBH, the latter feels "obsolete" nowadays. Maybe that's a problem.

I think it's really cool to create minimalist OSes. Something about perfboard projects and old, slow CPUs tickles my buttons. But I'm struggling to come up with a use case for this project in the End Times. In theory, this allows me to flash a ROM connected to an old through-hole, 8/16-bit CPU like a Z80, 8086, 6809 and 6502. I guess my issue is why and how would I do that during the end of the world?

I can't think of a way to come into possession of MPUs like that without intentionally buying them ahead of time. And if I'm going to stockpile those, I might as well stockpile a more capable MCU or MPU instead and flash it with something else. 99.9% of what I'd want to do with minimalist computers in the apocalyptic wasteland would be just fine without an OS. Bare-bones MCUs work spectacularly for control systems, wireless cryptosystems, data logging, etc.

Maybe I didn't look hard enough in the README [1], but I don't see how I'd bootstrap a system like this without already having a much more capable system on standby. Which comes back to the issue of... why?

[1] https://git.sr.ht/~vdupras/collapseos

Beyond LARPing the feasibility of building new computers post societal collapse, the website's proposal for collapse being so imminent is... weakly justified?

I'm not an optimist in today's societal climate, but the rationale [1] stems from peak oil and cultural bankruptcy. Peak oil is (was?) the concern of limited oil production in the face of static or increasing demand. The 2019 peak from big oil wasn't about declining production, but declining demand [2]. Which is good, if the ideal is long-lasting migration to renewables.

I won't try to predict the future w/r/t what current societal trends mean for the long-term success of global supply chains, but I would be greatly surprised if cultural bankruptcy alone causes the complete collapse of society in the next 5-10 years.

[1] http://collapseos.org/civ.html

[2] https://www.bp.com/content/dam/bp/business-sites/en/global/c...

Here's an ArXiv version: https://arxiv.org/abs/2502.19516

My drive-by opinion: "telecommuting" has an advantage in optics/marketing and flexibility over "work from home" for both business leaders and employees. If I tell a board of directors or shareholders that "80% of our workforce performs some fraction of their weekly tasks ______", I imagine the following:

- "from home" elicits images of relaxation and lost productivity, while "via telecommuting" sounds like a commute still takes place and work is just as productive

- "from home" sounds like retreating to a comfort zone, while "via telecommuting" sounds like embracing a new technology or skill

- "from home" sounds like remote workers ought to be performing their tasks from their domicile only, while "via telecommuting" sounds like remote workers can do their work wherever they are

If businesses had adopted "telecommuting" terminology, I believe business leaders would not feel obligated to push back in order to regain productivity. I think it's easier to attack the trend of WFH given the points above. I actually agree with the proposal that WFH is a weak terminology, but had never sat down and thought about it before.

From the perspective of initial adoption, I think it would have happened just as fast. Workers were thrust into remote work arrangements during COVID, and everyone would have quickly gotten the gist of what "telecommuting" means, so it would have been the new buzzword to attract talent in job listings just as "remote" or "work from home" have been. Just without the downsides in CEO perception.

I found on 136.0.0-ish that some settings persist despite checking/unchecking that box and restarting LibreWolf, but YMMV. I also manually inspect 'about:config' and search there for relevant settings (like 'fingerprint'). For fingerprinting, browser breakage is unlikely so toggling these hidden flags is easy.

Depending on one's threat model, any technique can be a secure strategy.

Is my threat model a network of dumb nodes doing automatic port scanning? Tucking a system on an obscure IPv6 address and never sharing the address may work OK. Running some bespoke, unauthenticated SSH-over-Carrier-Pigeon (SoCP) tunnel may be fine. The adversaries in the model are pretty dumb, so intrusion detection is also easy.

But if the threat model includes any well-motivated, intelligent adversary (disgruntled peer, NSA, evil ex-boyfriend), it will probably just annoy them. And as a bonus, for my trouble, it will be harder to maintain going forward.

No, no, you see: that would be a violation of local noise ordinance and is inhumane. Of course, a constant drip-feed of their banners and jingles in your private life is their right as a corporation.