amirhirsch

gotta charge less than the hourly cost of a B200 to remain employed

Captcha can detect the same person passing a captcha over and over. We shadow-ban to increase the cost of this kind of attack.

Source: I wrote the og detection system for hCaptcha

Yes. The moral failing of Saul (1 Samuel 15) is precisely not following the command to exterminate Amalek, but rather sparing Agag, and taking spoils when he was commanded to kill everything that breathes.

Being dismissive of the Bible is not as cool as you think it is: those who do not study the Bible are doomed to repeat it.

I understand the instinct to treat Bronze‑Age literature like fanciful fiction: engineers are wired to put "myth" in one bucket and "hard data" in another. But for better or worse, the Bible isn't just an ancient novel. It's the source code for a huge fraction of the world's legal systems, ethics, holidays, and political claims, including the one we are discussing. Dismissing it as "Harry Potter" misses the point:

If we're serious about reducing violence, we need to debug the real code people are running in their heads, not the straw‑man version.

The entire bible is a story of the Jewish people losing their land due to moral failings. Repeatedly. From Judges through 2 Kings, Israel repeatedly loses divine protection precisely because of its own wickedness. Being “chosen” meant bearing covenantal responsibility, not enjoying a blank moral check.

Deuteronomy 20:15-18 is more appropriate to the current conflict, as it relates to how the Jews should fight wars in the land of Israel. It commands the utter destruction of the inhabitants of the land, not sparing any that breathe (not just those who “pisseth-on-the-wall”)

Discussions about modern Israel/Palestine are full of shibboleths that reveal where people are drawing their information:

In the Hebrew the word in 20:16 for inheritance is “Nachala”. Worth Googling: Nachala is also the name of a present‑day Israeli settler movement led by Daniella Weiss, whose own literature says it’s “continuing the biblical mandate to settle the land.” In other words, the same term that the Torah uses for a gift that can be forfeited is now used as branding for a modern political project—illustrating how ancient vocabulary still shapes today’s arguments about the land.

For an example from the Palestinian side: you do not have a full understanding of Hamas if you do not know about the Hadith about the Gharqad tree. Hamas charter writers alluded to this story; many Palestinians learn it young, while most Israelis have never heard of it.

Recognizing these code‑words doesn’t require agreeing with the theology behind them. It simply keeps us from talking past each other—and, one hopes, from letting someone else’s apocalyptic script dictate who lives and dies. I think we all agree that the other-sides’ eschatology is a dumb reason to die.

I misremembered the story and checked on details: the arm broke in a fall while he was travelling several months prior to the move and it was Gerry joking that he punched a wall so he wouldn't have to move stuff.

Hi Keith! Another consideration was that the core computer programming skills needed to be taught more broadly to basically everyone and not just course 6 students.

Source: I shared an office with Keith next to Hal and Gerry--he taught me Java! Fun anecdote: we had to move RMS's stuff to the new CSAIL building because he had broken his arm punching a wall.

lol!! but also driving this is the fact is that each new thing speeds up the development of the next thing so 5x is more like (e^t)x

It started out as an innocent kv cache before the redis industrial complex became 5% of the GDP

great. now i have to spend my morning using vibe kanban to make a tui for vibe kanban with textual. i'll submit a pr as soon as it's finished.

i just realized the figure is showing the time breakdown as a percentage of total time, it would be more useful to show absolute time (hours) for those side-by-side comparisons since the implied hours would boost the AI bars height by 18%

Figure 6 which breaks-down the time spent doing different tasks is very informative -- it suggest: 15% less active coding 5% less testing, 8% less research and reading

4% more idle time 20% more AI interaction time

The 28% less coding/testing/research is why developers reported 20% less work. You might be spending 20% more time overall "working" while you are really idle 5% more time and feel like you've worked less because you were drinking coffee and eating a sandwich between waiting for the AI and reading AI output.

I think the AI skill-boost comes from having work flows that let you shave half that git-ops time, cut an extra 5% off coding, but cut the idle/waiting and do more prompting of parallel agents and a bit more testing then you really are a 2x dev.

MCP for Claude code that asks ChatGPT O3 for help. This is the way.

For each file in files: claude prompt with file

You can generally do map-reduce, also you can have separate git worktrees and have it work on all your tickets at the same time.

I built control computers for nuclear reactors. Those machines are not connected to a network and are guarded by multiple stages of men with automatic machine guns. It was designed to flawlessly run 3x boards each with triple-modular-redundant processors in FPGA fabric all nine processors instruction-synced with ECC down to the Registers (including cycling the three areas of programmable fabric on the FPGAs). They cycle and test each board every month.

What’s your source?

The whole Hacker News support thing gives good dev vibes but we had a support ticket for over-billing go unanswered for months (01288863) despite email pleas to a salesperson and then the ticket was closed without response. Made it impossible for me to convince my org to continue using Cloudflare.

No one will make long-term investments in US manufacturing capabilities in response to policy that will revert in 4 years.

Definitely missing from this list:

J. Cooley and J. Tukey, “An Algorithm for the Machine Calculation of Complex Fourier Series,” 1965

Put a 1 or 2 in front of that

But we're literally having this discussion with Gil Kalai in the next room.

The argument is that error correction doesn’t scale.

Welcome to Hacker News Gil! I’m a big fan of your work in complexity theory and have thought long and hard on the entropy-influence conjecture revisiting it again recently after Hao Huang’s marvelous proof of the sensitivity conjecture.

To answer your question on why the hyped fantastic claim, as you must know, the people who provide the funds for quantum computing research almost certainly do not understand the research they are funding, and need as feedback a steady stream of fantastic “breakthroughs” to justify writing the checks.

This has made QC research ripe for applied physicists who are skilled in the art of bullshitting about Hilbert Spaces. While I don’t doubt the integrity of a plurality of the scientists involved, I can say with certainty that approximately all of the people working on Quantum Computing Research would not take me up on my bet of $2048 that RSA2048 will not be factored by 2048 —- and would happily accept $204,800,000 to make arrays of quantum-related artifacts. Investors require breakthroughs or the physicists will lose their budget for liquid gases — certainly exceeding $2048.

While there might be interesting science discovered along the way, I think of QC a little like alchemy: the promise of unlimited gold attracted both bullshitters and serious physicists (Newton included) for centuries, but the physical laws eventually emerged that it is not scalable to turn lead into gold. Similarly it would be useful to determine the scaling laws for Quantum Computers. How big of an RSA key is needed before even a QC exceeds the total number of particles in the universe to factor it in reasonable time? Is 2048 good enough that we can shelf all the peripheral number-theory research in post-quantum-cryptography? Let’s not forget the mathematicians too!

Yes. There are many Jewish lending organizations designed around the principle that loaning money to establish a new business is the highest form of charity (see Maimonides 8 levels of giving). For example, there is a 127-year-old organization based in San Francisco that offers loans up to $50,000. The expectation is once someone establishes themselves they will contribute back to the lending organization. The workaround to halachic rules of interest that is applied in Israeli banking is equity co-ownership where the lender is paid back for their ownership shares at an established interest rate; the term for this is called “heter iska.”

You are not logic-ing correctly. The financiers were Jewish. Jews do not charge interest to other Jews; if Columbus were not Jewish they would have charged interest. This is written in Deuteronomy 23:20-21 as well as Leviticus and Exodus and also discussed by Maimonides

Another indication that Columbus was part of the Jewish community is that the Jewish financiers Luis de Santángel and Gabriel Sánchez advanced interest-free loans to finance the journey.

YC benefits strongly from network effects; the value for each founder grows superlinearly with more founders. Grow faster!

Maximum performance is only one corner for optimization. We do not operate chips at high temperatures in our pocket; the goal is to get low power for the same functionality to have a long battery life, in which case parallelization means doubling the number of transistors while halving clock rates lowering the total power quadratically. It is also not more expensive per area to stack logic: by lowering the area of each die, you increase the yield of each in the stack.

Weather models are chaotic, are ML methods more numerically stable than a physics based simulation? And how do they compare in terms of compute requirements? the Aurora paper seemed to be promising, but I would love a summary of comparison better than what I get out of Claude.

Once upon a time I converted spectral-transform-shallow-water-model (STSWM or parallelized as PSTSWM) from FORTRAN to Verilog. I believe this is the spectral-transform method we have run for the last 30 years to do forecasting. The forecasting would be ~20% different results for 10-day predictions if we truncated each operation to FP64 instead of Intel's FP80.

https://scottaaronson.blog/?p=8310 suggests perhaps the next decade for useful quantum computer. this suggests the scale for RSA2048: https://arxiv.org/pdf/1905.09749

The thing about this bet is you don't have to pay until 2048 whereas I have to pay as soon as it happens and accounting for savings rates the bet will cost about a dime a day.

I will bet $2048 that a quantum computer will not factor RSA2048 by 2048.