bumbledraven

https://archive.ph/NLgU9

TFA links to https://joshcollinsworth.com/blog/tailwind-is-smart-steering, which is about Tailwind, but makes multiple distinctions and points that could just as well apply to LLMs, e.g.:

Builders value getting the work done as quickly and efficiently as possible. They are making something—likely something with parts beyond the frontend—and are often eager to see it through to completion. This means Builders may prize that initial execution over other long-term factors.

Crafters are more likely to value long-term factors like ease of maintainability, legibility, and accessibility, and may not consider the project finished until those have also been accounted for.

In my view, the more you optimize for building quickly, the more you optimize for homogeneity.

800-word response by https://en.wikipedia.org/wiki/Mark_Lucovsky, who was on the team that built Windows NT: https://nitter.net/marklucovsky/status/2052828852490285390#m.

It begins:

In David Crawshaw’s recent post “The agent principal-agent problem” there’s a lot of insight beneath the headline “Code review is broken.” Worth reading carefully.

Toward the end, David reflects on what he calls the old “cowboy” development culture at Microsoft in the 80s/90s. Not much has been written about that era, mostly because there was no social media, no laptops everywhere, no phones recording daily engineering life.

A few thoughts from someone who lived it.

Conclusion:

A lot of software quality did not come from pre-commit review gates.

It came from tight teams, deep ownership, brutal integration pressure, system-wide stress, and developers who fully understood the machinery they were standing on.

This article has one neat Google search trick I hadn't known about: `AROUND(#)`. But I am skeptical of much of the rest of it.

Searching for `“can anyone recommend”` to get unfiltered recommendations is an interesting hack, but I feel it's not too reliable. At reddit, you could ask this, and an unknown percentage of responses could be shills or bots.

I'm also skeptical of how much the suggested `@reddit` differs from just `reddit`. The description says it's for social media handles, but reddit is a platform, not an individual user's handle. I suspect google looks for the user's intent to see results from a particular site or social media platform and uses that signal to influence the ranking, and I doubt '@' has much of an effect on that process.

The results Google omits tend to be less trafficked and less search-optimized, which frequently means they’re more substantive and written for readers rather than algorithms

Really? I call BS. Every time I've looked at the omitted results they've been very similar to ones I've already seen.

I stopped reading after that.

Details on the strategy employed by the human are at https://goattack.far.ai:

We discovered simple adversarial strategies that beat superhuman Go AIs, and find that adding defenses helps but does not eliminate the problem. Our cyclic adversary beats the state-of-the-art KataGo AI more than 97% of the time at superhuman settings. This strategy is simple enough to be replicated by an amateur human player and transfers to other superhuman Go AIs.

https://archive.is/EBcVa

A design doc with a robust "alternatives considered" section.

It would be neat if a fuzzer could help set a new tool-assisted speedrun (TAS) record.

You fill Harvard of Stanford with only people with 1600 SATs will turn them into places you dont really want to go to.

Isn’t that basically Caltech? They had a 3% acceptance rate in 2023, the lowest in the nation. https://www.usnews.com/best-colleges/rankings/lowest-accepta...

Do you have a citation for this? The most Gemini could say is: "While research has not identified a specific tsunami stone located at the Fukushima Daiichi site that was directly violated, the spirit of these ancient warnings was undeniably ignored." (https://aistudio.google.com/app/prompts?state=%7B%22ids%22:%...)

Absolutely no causal link was shown. Maybe the nerve pains for which doctors prescribe gabapentin increase the risk of dementia on their own, or maybe there is some third factor that causes both nerve pain and dementia.

What model were they using?

Yes they do, but keep in mind the distinction between "armies" and "irregulars". The first two sentences in the article are:

Arabic-speaking armies have been generally ineffective in the modern era. Egyptian regular forces did poorly against Yemeni irregulars in the 1960s.

I have been writing the same thing by (ab)using the existing unit of measurement known as a bel (B), which is most commonly seen with the SI prefix “deci” (d) as dB or decibel. I write the speed of light as 8.5 Bm/s (“8.5 bel meters per second”), which resembles the expression 20 dBV (“20 decibel volts”) shown at https://en.wikipedia.org/wiki/Decibel.

Full interview: https://www.youtube.com/live/esCSpbDPJik

Translation via Gemini (2.0 Experimental Advanced)

Kidnapping of One of the Co-founders of the Startup Ledger: Major Investigation in the Cher Region

As part of the investigation into the disappearance of a businessman, numerous gendarmes and police officers were deployed in the Vierzon area on Wednesday, January 22nd. According to our information, they are investigating the kidnapping of one of the co-founders of the startup Ledger, David Balland.

By Antoine Bailleron and Geoffroy Jeay

Published on January 22, 2025 at 11:00 PM

David Balland created, in Vierzon, in 2014, with four other partners, Ledger, a startup specializing in ultra-secure cryptocurrency wallets, notably for Bitcoin.

The significant deployment of gendarmes in a street in Méreau on Tuesday, which surprised the local population, moved to the Vierzon area on Wednesday. Numerous gendarmes and police officers were visible on the roads leading to Vierzon, and even in the sub-prefecture of Cher. A gendarmerie helicopter was seen at low altitude. In the afternoon, gendarmes even intervened on the Quai de l'Yèvre in Vierzon, breaking a window and the door of an apartment.

[image]

Gendarmes intervened on the Quai de l'Yèvre in Vierzon, breaking a window and the door of an apartment.

According to our information, they are investigating the disappearance of a businessman, one of the co-founders of the startup Ledger, David Balland. The man was reportedly kidnapped. With a view to a ransom demand? Nothing has been revealed.

as we’ll discuss, it’s a basic metamathematical fact that out of all possible theorems almost none have short proofs

Where in the article is this discussed?

"More human than human" is our motto. https://youtu.be/ZbgmYhqFO-4?t=30

I’m astonished that Verizon didn’t charge separately for installation. Seems like it would take many years months of service to recoup the cost.

While defining x/0 = 0 is sound, it has tradeoffs, as OP acknowledges. In "Papers with computer-checked proofs" (2024), D. J. Bernstien wrote:

There is a claim ... that “dividing by 0 is not allowed in mathematics, and hence this cannot be relevant to their work”, but in fact a mathematician is permitted to deduce b ≠ 0 from c = a/b, since a/b is undefined for b = 0. Redefining the notation to allow b = 0 breaks this. Unless it occurs to the author to state a conclusion c = a/b and a conclusion b ≠ 0, the proof assistant won’t check that b ≠ 0, whereas the reader will think that this has been checked.

https://cr.yp.to/papers/pwccp-20240727.pdf

https://x.com/esrtweet/status/1859556171918479626 (2024-11-21):

Evil has a sound. It's the sound of humans screaming in agony as they die. Aztecs made implements to simulate this sound. They have been found in the hands of sacrificial victims… the Aztecs considered the ceremonial experience incomplete without the screaming, and built death whistles to simulate the proper sound of agony.

Only if they have — or can reasonably develop — the means to produce cutting-edge chips on their own, which is (arguably) not the case.

In the Gregorian calendar, January 1 falls on a Saturday in any year that is a multiple of 400. (This follows from any correct day-of-the-week algorithm, such as https://firstsundaydoomsday.blogspot.com/2009/12/quick-start....) Since each 400-year cycle begins on the same day, rather than balancing out, any imbalances would actually accumulate.

To check, see this 5-line brute-force calculation in JavaScript: https://news.ycombinator.com/item?id=42024474

It turns out that New Year's Day is a bit more likely to fall on Sunday than Saturday. To confirm, here's a bit of simple JavaScript that checks a complete 400-year cycle. You can run it in your browser's console:

  const counts = Array(7).fill(0);
  for (let year = 2000; year < 2400; year++) {
    counts[new Date(year, 0, 1).getDay()]++;
  }
  console.log(`Number of Saturdays: ${counts[6]}; Number of Sundays: ${counts[0]}`);

  Number of Saturdays: 56; Number of Sundays: 58

I don't see it there: https://archive.org/details/boris-a.-kordemsky-the-moscow-pu...

Do you recall the solution? It seems like a tricky problem. No approach occurs to me aside from a brute force analysis of the 400 year Gregorian calendar cycle, accounting for the complete leap year rules (under which, e.g., 1900 was not a leap year but 2000 was).

+1, that fixes the voice recognition part, but it doesn't automatically read Claude's response aloud.

I wish I knew of a good way to voice chat with Claude, and unfortunately, the new mobile app isn’t it. The voice recognition is terrible. Does it just use iOS’s built-in text-to-speech? It needs to be as good as Whisper to be useful.

But at the end of the day he needs to make a concrete prediction that differs than the current view in order to have people devote a lot of time studying his world view

Even if it doesn't make any different concrete predictions, a new way of thinking about things can attract scientists' attention. The Many Worlds interpetation of QM is an example.

Kudos to htmx.org for hosting this essay.

Not if compression is involved.