thornewolf

We have a bad universal classifier now (via Jev). 0->1, one might say.

A bad universal classifier does suggest a good one later. And that is exactly what I would call "theoretically doable"

That said, I don't think that Jev is a magic breakthrough or anything. I think it is just a particularly good narrative with an easy way to try it out.

Write a prompt, evaluate the prompt, understand that is succeeds 95% of the time.

Write a new prompt, evaluate, it now succeeds 99% of the time. Measure what changes between prompt #1 and prompt #2, understand what contributed to the performance jump.

Write a third prompt, this one succeeds 100% of the time. Increase the size of your evaluation set, find a 1/5000 error-class and a 1/10000 error-class, add some explicit code to correct for this cases.

Roll out to production, collecting usage metrics. You make some tweaks to your harness, your prompts. Eventually you have confidence that your system has fewer mistakes than 1 in 100k.

Now, multiply this iteration across all your different prompts and different ways that they might interact with one another.

without getting too far into it i would just like to note that i am in both these circles

some initial function like

  v = setup()
  if v == 1:
    side_effect_1()
  elif v > 1:
    side_effect_1()
    side_effect_2(v)
  else:
    raise Exception()

then we can "refactor"

  v = setup()
  if v < 1:
    raise Exception()
  
  side_effect_1()
  if v > 1:
    side_effect_2(v)

i know that this might seem "dumb" that the code was ever setup the first way but code can grow into that shape pretty easily. this refactor "removes" the v==1 branch. this new code also follows the "early return" pattern, which improves readability.

to add some context from "the other side": last year, i listened to 2930 distinct albums of which 50% were albums I had not listened to before. Even on a 30 year timeframe and a good discount factor for album repetition I don't think I can fit into the budget you propose without making significant concessions in my discovery process.

My response is going to be about Gemini generally and less about NotebookLM.

Google's last frontier model release was Gemini 3.1 Pro, which was in February of this year[1]. At the time, it was ahead of the (at the time) flagship models of Opus 4.6 and GPT 5.2/5.3. From my recollection of the time, it was the best model in the world.

Anthropic released Opus 4.5 Nov '25, 4.6 in Feb '26, 4.7 in April, 4.8 in (late) May. Then Fable in June. 4.7 beat 3.1 Pro on multiple metrics. Fable eats it for breakfast. However, I want to note the 3 month gap between those first two Opus versions.

OpenAI released 5.2 Dec '25, 5.3 Codex Feb '26, 5.3 Instant Mar, 5.4 Mar, 5.5 (late) May, 5.6 July. 5.4 beats 3.1 Pro on agentic benchmarks[2], seems to be similar/losing on non-agentic. 5.5 seems stronger than 3.1 Pro[3].

Gemini 3.5 Pro is alleged to be launching within the week. Why do I type this all out? Because I think Google is getting a bad rap. They are delayed on a frontier release by a month or two and are being regarded as if they cannot release frontier models. I think their last release demonstrates strength and we need to see a weak release before we call them "behind" (in any reasonable sense). These companies swap back and forth constantly. I recall a multi-month span where 2.5 Pro was just the best thing out there by a large margin (in my opinion).

[1]: https://blog.google/innovation-and-ai/models-and-research/ge...

[2]: https://www.anthropic.com/news/claude-opus-4-7

[3]: https://www.anthropic.com/news/claude-opus-4-8

just read this when reviewing OpenAI's "spinning up" documentation as it was linked there!

The author explain this in more detail in the article; I also was initially confused. The key details come from their broken down pandoc command. Specifically, these two flags:

  --pdf-engine=typst
  -V template=article.typ

Despite the input content being a .md file, we see that the .md contents populate the .typ file. Pandoc then understands how to convert the populated .typ file into a .pdf. The author also notes in their hyperlinked .typ document where the body content is placed:

  // THIS IS THE ACTUAL BODY:

What is not clear to me is the actual mechanism that tells pandoc where to place the body contents in the template. I presume it's some "magic" from the pandoc docs. The pandoc templates documentation[1] does reveal to us that `-V` is setting a variable called `template`. I don't have pandoc locally, but from another post[2] from the same author and a chatgpt query, it seems that the `body` symbol in the template is the chosen substitution symbol.

[1]: https://pandoc.org/demo/example33/6-templates.html

[2]: https://imaginarytext.ca/posts/2024/pandoc-typst-tutorial/

not dead, they are on "luma". quality is higher in some ways, lower in some ways when compared to "old school" meetups

I can't smell much at all. one time when i was 17 my friend told me i was kinda smelly (i had just exercised). I've been stressed about it since. the human brain can latch on to the oddest things

there is something "wrong" with the ux that is hard to pin down. these things generate even text summaries more rapidly than i can read them. i need a better method for dumping info into my brain + dynamic control (if necessary)

They have named themselves sqliteai and are releasing a library called adam. this is just advanced seo. seems neat though

it's a security feature. otherwise my malicious site could check for cdn.sensitivephotoswebsite.com and blackmail you if it was cached already

more ai articles on my front page :(

I noticed this article is written by AI. Have you considered adding a disclosure?

just disclose please. i churn on non-disclosed ai generated content as a matter of principle. i don't necessarily churn on other variants

this article comes back 100% ai (for the opening). i will not reveal the tell that made me churn

I have swapped to using gemini over chatgpt for casual conversation and question answering. there are some lacking features in the app but i get faster and more intelligent responses.

For the comment from "Is Show HN dead? No, but it's drowning", the article author editorializes

"One of the great benefits of AI tools, is they allow anyone to build stuff... even if they have no ideas or knowledge.

One of the great drawbacks of AI tools, is they allow anyone to build stuff... even if they have no ideas or knowledge."

into

"One of the great benefits of AI tools is they allow anyone to build stuff, even if they have no ideas or knowledge. One of the great drawbacks is they allow anyone to build stuff."

which removes the rhetorical effectiveness of the comment (and also breaks the promise of a quotation). I recommend that OP represents the source exactly.

____

I now see that this article contains multiple GPT-isms

dispell the hate from your heart

I'm not the article author

it's because this argument of "what is $0.01 more?" can be extended forever, implying you are willing to pay an infinite amount of money for anything. since we know this is silly, we try to understand what our "real" maximum is. this is difficult to do for exactly the reasons you mention in your comment! surely $0.01 is negligible! there is a tension here.

and so, absolute max price is not a fantasy - the world would be absurd if it were - but instead its a real and difficult to construct value

as someone who worked at the company, i understood the meaning behind the tweet without the additional clarification. i think she assumed too much shared context when making the tweet

while we can't trust their word as absolute truth, they did specifically say they still not do this in the article

I estimated that i was 1.2x when we only had tab completion models. 1.5x would be too modest. I've done plenty of ~6-8 hour tasks in ~1-2 hours using llms.

some forms of meditation can be. it's a very general term

I started typing in no caps in most conversation around 2013-2016, when I first started playing League of Legends and using Discord. Other people from this same age bracket as me have similar experiences (if they were "chronically" online) and similar behavior patterns. It's not pretentious imo.

maybe so, though an inaccurate claim. the ai is the value add here (and quite a value add based on the other comments in the thread). we typically reserve the word "slop" for ai generated content that is of low quality or no value add. this website seems to be both of quality and value ad and it would be difficult to argue otherwise.

This comment is likely AI. Consider author post history as well.

(this comment)

> This idea sits in a really interesting space because on paper

(previous comments)

> I really like this class of work because it sits at a strange intersection:

> It’s wild how Voyager forces two truths to sit together:

The pattern is "<compliment intellectual stimulation> <make note of juxtaposition>"

> becomes workflow-aware, not just price-aware.

> human mental model instead of the mathematically convenient one.

> run for that mindset more than for the tech.

My meta comment is not breaking the HN guidelines by letter, but may be spiritually breaking guidelines. See:

> Please don't post insinuations about astroturfing, shilling, brigading, foreign agents, and the like.

I think clarification on AI-accusations should be added to the guidelines if it falls under this class of comment.