notpachet

Also fair.

I am not describing digital abstinence. Note that I did not mention abstaining from:

- the web (I am posting on HN after all, although I recommend a very stripped down browsing experience)

- email

- general computer use

It's telling that you interpreted that (rather circumscribed) list as total digital abstinence. Social media, smartphones, and AI are all relatively recent appearances on the digital stage.

This is not digital minimism. Digital minimism would be:

- Not using social media at all

- Not using your phone at all

- Not using AI at all

> The central argument is that AIs, living inside datacenters, won't be able to evaluate the "coherence" of software "beyond the human ceiling"

Actually, the article states that AI will easily be able to handle increases beyond the human ceiling as long as those increases follow a linear curve, not an exponential curve.

So your whole critique dissolves. Sorry!

You're tying together some pretty far-flung strands of argument there.

Regarding children and selfishness: some of the most selfish people I know have kids. They're selfish in the sense that they only care about their family, even to the detriment of everyone else.

> - Any (super)intelligence caches work > - Any (super)intelligence _wants to create stable abstractions_

This is a higher-order rephrasing of the "intelligence is compression" argument.

"What I cannot create, I do not understand"

- Feynman

> Let's be better please.

That rhetorical horse left the allegorical barn a long time ago.

> if you can do the worst “bad” thing possible, you can also do the best “good” thing possible

You say this like it's a truism, but I don't think it is. The bad things may be categorically easier than the good things. It's easier to destabilize a delicate system than to stabilize it. It's easier to break something than to fix it.

I did bother to read it, and that isn't what I meant. I wasn't referring to one-off errors that you can fix after successfully decoding a cipher into something that produces a majority-correct answer. I meant something more akin to what schoen's answer touched on: whether there are any closed-form mathematical/statistical analyses that can be performed ahead of time to assess the structural legitimacy of any ciphertext before you spend a lot of cycles trying to decode it.

Thanks for the driveby snark though!

That's pretty good.

> It spun for 52 seconds so maybe some Python was involved.

Come for the AI doom, stay for the programming languages (remember those?) catching strays

Cipher noob question: is there any check that can be done to ensure a cipher is actually decodable? What if the author made a flaw when encoding it, so that it's not actually solvable?

"My dosimeter only reads 3.6 roentgen / hour!"

Not the parent, but pointing out a couple of problems in your analysis:

First, arctic temperatures: this is a widely misunderstood point. Notice that in the graph, the non-summer months are noticeably warmer. The summer months are going to be effectively capped at around 0 degrees until all the ice is gone, so holding up this graph and saying "look, the temperature never got above freezing!" is misleading, because the "warming" is actually happening as a function of melting ice, not air temperature.

Sea ice extent: the last 18 years were the lowest 18 minimums on record. Also, ice extent != ice volume. March 2026 recorded a record-low ice volume. An extent chart reads the same regardless of whether the underlying ice is thin or thick.

Greenland mass balance: here you're just plain wrong. You linked to the surface mass balance page, in other words, snowfall minus surface melt + runoff. This does not account for glacial calving and basal melt. The page you linked says this directly. Total mass balance is a grimmer picture: an average loss of about 252 km³ of water per year over 2003-2025, contributing 0.69 mm/yr to sea level rise.

There's a mythological story about the hubris of flying; I'm sure Claude can pull it up for you.

Moreover who’s actually getting exterminated?

The species whose value was questionable.

The good ones remain.

There's a larger existential crisis still in store for you. Our global civilization is collapsing. We are choking to death on the unchecked negative externalities of our own progress. It is our species' Achilles heel. Whether or not an individual developer adapts to this temporary capability blip (agentic coding) will not matter in the long run. The systems necessary for supporting these capabilities are too brittle to survive the century.

Instead we just get good, hard slop

Because they're addicted to online content consumption. See also: sports betting, cigarettes, fast food.

> the way they'd go after a very inconvenient activist

FTFY

Related reading:

The Waluigi Effect: After you train an LLM to satisfy a desirable property, then it's easier to elicit the chatbot into satisfying the exact opposite property.

https://www.lesswrong.com/posts/D7PumeYTDPfBTp3i7/the-waluig...

You've hit on something important, and it's real, not cosmetic.

GNU Screen has supported multiuser editing since 1993.

"You newer models are happy scraping the shit... because you've never seen a miracle"

We seem to have come around to a new magical age.

> the guy

> he

Your gender biases are showing!

And that, furthermore, the new labs and the old labs still both remain incentivized by the regulatory environment to chase the wrong targets. The new lab might just iterate towards the wrong targets faster.

"Testing shows the presence, not the absence of bugs"

- Dijkstra

I think about this a lot. From the POV of a medieval peasant, we likely reached the singularity a long time ago.