Hacker News: aesthesia

New comment by aesthesia in "Changes in the system prompt between Claude Opus 4.6 and 4.7"

aesthesia — Mon, 20 Apr 2026 02:04:58 +0000

Leaked/extracted system prompts for other chat models, particularly ChatGPT, are often around this size. Here's GPT-5.4: https://github.com/asgeirtj/system_prompts_leaks/blob/main/O...

New comment by aesthesia in "Slop Cop"

aesthesia — Fri, 17 Apr 2026 23:56:49 +0000

I think this would come off a lot better if the recommendations weren't so absolute. I like the effect of a multicolored slab of highlights calling out every LLM cliche in a passage. Yes, the slop style is not just the sum of these individual patterns, but they're definitely significant contributors to the effect, and they're worth being aware of in your own writing regardless of their association with LLMs. You just can't treat it as a list of must-resolve errors (same as with any writing feedback, really).

New comment by aesthesia in "Playdate’s handheld changed how Duke University teaches game design"

aesthesia — Fri, 17 Apr 2026 16:23:24 +0000

At least in some fields, advanced courses are the most likely to have lower cost textbooks. Real analysis textbooks are usually cheaper than calculus textbooks. It's the introductory courses that tend to have $200 behemoths attached to online homework platforms optimized for ease of grading rather than student learning.

New comment by aesthesia in "Claude Opus 4.7"

aesthesia — Thu, 16 Apr 2026 15:51:17 +0000

The new tokenizer is interesting, but it definitely is possible to adapt a base model to a new tokenizer without too much additional training, especially if you're distilling from a model that uses the new tokenizer. (see, e.g., https://openreview.net/pdf?id=DxKP2E0xK2).

New comment by aesthesia in "Qwen3.6-35B-A3B: Agentic coding power, now open to all"

aesthesia — Thu, 16 Apr 2026 15:32:57 +0000

> Anthropic has also admitted that the bugs found by Mythos had not been found by using a prompt like "find the bugs", but by running many times Mythos on each file with increasingly more specific prompts, until the final run that requested only a confirmation of the bug, not searching for it.

Unless there's been more information since their original post (https://red.anthropic.com/2026/mythos-preview/), this is a misleading description of the scaffold. The process was:

- provide a container with running software and its source code

- prompt Mythos to prioritize source files based on the likelihood they contain vulnerabilities

- use this prioritization to prompt parallel agents to look for and verify vulnerabilities, focusing on but not limited to a single seed file

- as a final validation step, have another instance evaluate the validity and interestingness of the resulting bug reports

This amounts to at most three invocations of the model for each file, once for prioritization, once for the main vulnerability run, and once for the final check. The prompts only became more specific as a result of information the model itself produced, not any external process injecting additional information.

New comment by aesthesia in "Good conversations have lots of doorknobs"

aesthesia — Sun, 21 Aug 2022 19:09:55 +0000

This reminds me of the “no hello” proposal for workplace chat messages (e.g. https://nohello.net/). It’s much less of a big deal for personal communication, but I can understand wanting someone to just say what they want to say without a manual SYN/ACK first.

New comment by aesthesia in "The Secretary Problem"

aesthesia — Fri, 19 Aug 2022 12:57:14 +0000

Interestingly, the traditional algorithmic solution is inherently asymmetric: it gives better outcomes to one gender than the other.

New comment by aesthesia in "The Secretary Problem"

aesthesia — Fri, 19 Aug 2022 12:51:13 +0000

Crucially, the optimal solution depends on the specific constraints. So you can’t just take a solution derived under one set of assumptions and use it in another situation without losing whatever guarantees it made in the original case.

New comment by aesthesia in "Your price at the pump went up. So did Saudi Aramco's profits – to a new record"

aesthesia — Mon, 15 Aug 2022 00:34:48 +0000

How is the market rate determined in those places? If no one can lower prices unilaterally, it seems like price discovery just can’t happen.

New comment by aesthesia in "The mitochondrial NAD+ transporter SLC25A51 is a fasting-induced gene"

aesthesia — Mon, 08 Aug 2022 00:20:46 +0000

In mice, of course. What are the broader implications of this supposed to be?

A Mormon housewife turned a fake diary into an enormous best-seller

aesthesia — Mon, 01 Aug 2022 01:37:43 +0000

Article URL: https://www.newyorker.com/magazine/2022/08/01/how-a-mormon-housewife-turned-a-fake-diary-into-an-enormous-best-seller

Comments URL: https://news.ycombinator.com/item?id=32300864

Points: 94

# Comments: 89

New comment by aesthesia in "As TikTok grows, so does suspicion"

aesthesia — Sun, 10 Jul 2022 17:35:21 +0000

Authoritarianism and military aggression are not the same thing.

New comment by aesthesia in "No Refrigerant Left Behind"

aesthesia — Fri, 01 Jul 2022 20:26:50 +0000

> So long as we are burning natural gas for energy, it's better to use it directly in your home, vs have someone else burn it, make electricity, then use that.

Only if you're using resistive heating. Heat pumps run on natural-gas-produced electricity can be at least as efficient as direct natural gas combustion for heating, and they automatically transition to cleaner sources of energy as the grid does.

> you would have to use reusable ones hundreds of times, and never wash them - ever, for them to be better

I understand that this is the case for cotton bags, IIRC due to high water use in cotton production, but for other types of reusable bags the threshold is lower.

> people reuse around half of them for garbage bags

This estimate seems like it's significantly too high. I do most of my grocery shopping at places that don't provide free plastic bags, and yet I still end up with far more single-use plastic bags than I could ever use for garbage. I would guess that no more than 10% of single-use bags actually get reused for trash.

New comment by aesthesia in "Twitter board adopts poison pill after Musk’s $43B bid to buy company"

aesthesia — Sat, 16 Apr 2022 14:00:19 +0000

Uh, wouldn't spammers just...stop using lots of hashtags? Part of what makes the problem difficult is that spammers are agents who respond to the techniques you use to stop them.

New comment by aesthesia in "Why Does Windows Use Backslash as Path Separator? (2019)"

aesthesia — Tue, 01 Mar 2022 04:42:48 +0000

Cutler’s Wikipedia page puts his move to Microsoft in 1988, not 98, which makes the timeline clearer. He was directly involved with the creation of the Windows NT project, which started well before 1998.

New comment by aesthesia in "Graph Theory and Linear Algebra [pdf]"

aesthesia — Thu, 24 Feb 2022 23:24:48 +0000

I don't think that follows. Since the adjacency matrix contains O(n^2) _bits_ of information, the number of adjacency matrices of graphs is O(2^(n^2)). This makes sense: graphs with vertex labels {1,...,n} are in bijection with symmetric n x n binary matrices with zero diagonal. It's difficult to quantify how much information is lost when we reduce this matrix to its spectrum, since the eigenvalues are real numbers, but there are O(n) degrees of freedom, and in principle plenty of room to distinguish any two graphs.

Of course, as a sibling comment notes, unlabeled graphs are often more interesting, for which one has to look at equivalence classes of adjacency matrices. The spectrum is nice here because it is automatically invariant to vertex permutations and hence a graph isomorphism invariant. The existence of nonisomorphic cospectral graphs is not immediately obvious, although there are simple examples, and that there are infinite families of cospectral pairs is even less obvious. So at the very least, it's interesting that the spectrum is not a complete invariant, and that it does work well for certain classes of graphs.

New comment by aesthesia in "Swiss back further restrictions on tobacco advertising"

aesthesia — Mon, 14 Feb 2022 04:19:04 +0000

Yes, you should drink when you’re thirsty, and forcing yourself to drink some arbitrary quantity of water every day is probably pointless. But water without any dissolved minerals is also bad for you, and many bottled brands add other salts (eg calcium chloride) to filtered water for flavor. I’ve also never experienced the phenomenon you describe where water with sodium fluoride fails to satisfy my thirst. While I certainly prefer the flavor of filtered water, drinking tap water doesn’t make me crave any other drink. Has there been any research done on this topic?

New comment by aesthesia in "What kind of Apple Mac did Arthur Dent have?"

aesthesia — Sat, 12 Feb 2022 15:54:05 +0000

Not quite: at one point there's a reference to a non-wizard character having a PlayStation, which I found kind of jarring to read at the time, precisely because it pinned the story down to a very particular point in time.

New comment by aesthesia in "Theses on Sleep"

aesthesia — Thu, 10 Feb 2022 23:57:00 +0000

Connect an LED directly to a battery (as in many cheap flashlights) and there will be no flicker.

New comment by aesthesia in "Can a biologist fix a radio?–Or, what I learned while studying apoptosis (2002) [pdf]"

aesthesia — Fri, 28 Jan 2022 23:55:02 +0000

This paper inspired another fun followup: "Could a Neuroscientist Understand a Microprocessor?" [0] The authors investigated a 6502 running Atari games using popular neuroscience methods, finding things like transistors that are uniquely necessary to run Space Invaders, but, of course, never getting close to an actual understanding of how the processor functions. It really highlights how even with the huge amounts of data we are able to get from biological systems now, there's still a lot of information we paradigmatically can't understand.

[0]: https://journals.plos.org/ploscompbiol/article?id=10.1371/jo...