HNHacker News
TopNewBestAskShowJobs

blagie

5,691 karma · joined March 22, 2018

submissionscomments
blagie··on Perforce charges $500 for training training videos.. and it's AI narrated
> However, I continue to be amazed that Perforce survived Git. I assume its ability to handle large binaries has been what saved it.

Curiously enough, the git data model is ideal for handling large binaries. The place it crashes-and-burns is the user space.

blagie··on LLM-Integrated Multivariable Calculus Course
> Everything I do is driven by a single goal: to make my startup successful so I can innovate in edtech. Nothing else.

I believe you entirely. That comes across in everything you say. Huge harm has been done in education precisely by people with similarly amoral motivation. History of education is filled with case after case of amoral motivations leading to immoral behavior.

Other industries have this more extreme. Fritz Haber is a biography worth reading.

The comments you're making aren't very honest -- I suspect with yourself more than with anyone else. For example:

> Maintaining an open-source project requires significant time and effort.

No. It doesn't. You need to flip the "public" bit on github and post a "caveat emptor" shingle. The effort you're cloning -- manim -- did exactly that. The result was a synergistic, friendly, manim-ce.

And the author has been able to work full-time in this space, his entire career, without doing anything evil, unethical, harmful, or unpleasant, and strictly advancing the world.

You're competing -- rather than cooperating -- with folks like him.

> I simply need enough protected space to innovate competitively and give the company the best chance to succeed.

You don't seem to understand how this works. A few points:

- Education is a highly-regulated industry for good reasons (student harm, societal harm, privacy risks, etc.), and education research goes through IRBs for similar reasons.

- Startups can work on IP, but usually -- and especially in education -- networks, connections, support, etc. are more important. It's hard to build an IP moat.

- Startups have a 95% failure rate. I'd give this one more than that.

A few corollaries:

a) With the current model, you'll have a hard time building partnerships with the people you need, and indeed, people (myself included!) will be rooting for you to fail. That weakens, not strengthens, your odds of success.

b) An example of another way to accomplish your stated goal is a five-year research grant, or a position which supports doing this kind of work (e.g. instructional staff at your university). Those have >> 5% success rates.

c) Open business models aren't less likely to be successful here than closed ones. Red Hat sold for $34B.

Otherwise, my best wish for you is to fail fast, and to learn a lot from the process.

blagie··on LLM-Integrated Multivariable Calculus Course
I did try it.

I'd provide feedback on how to make this better, but to be frank, I don't want to see this made better. I'd like open education made better. I'm on the other side here. Even if I did, I'm pretty sure you wouldn't listen.

In my career, I've done both open and closed models (more of the former in recently, more of the latter early career). I'm open to both, but education should be open, for a whole slew of reasons I can enumerate.

blagie··on LLM-Integrated Multivariable Calculus Course
> All we want to do is innovate.

That's very clear. That's why we have spyware, ad tech, WWII-era German chemical weapons, the East India Company, Oracle's business model, fintech, ...

Ask anyone 40 years older. Regrets focus on time with family and meaningful work. Few people think: "Did I innovate enough?"

> We simply can't open source this because we're rewriting and rebuilding it every day.

I don't quite see the logical connection in this sentence.

Everything I've worked on in the past 15 years is open-source. I suspect you'd:

- Be able to find modest amounts of funding to do this in the open space (e.g. within the academy, as a teaching professor, adjunct, or similar) -- enough to sustain yourself, but without a promise of riches

- Be able to "innovate," make more E($), and do less harm in corporate America

blagie··on LLM-Integrated Multivariable Calculus Course
LLM-written clones of free software into proprietary seem to be a growing trend now.

With manim, it's probably legal (I think MIT license?), but in general, it makes the world a worse place.

blagie··on LLM-Integrated Multivariable Calculus Course
This sort of thing makes me depressed. The videos are LLM slop. You're a confused student. You do a web search. You run into something like this. It looks high-quality and professional. It is dogshit which no one can understand. The result is a confused learner who feels bad about themselves.

At the same time, thousands of low-quality resources overwhelm good content.

Integrating with an existing resources like MOOCulus could add a ton of value, in contrast, but wouldn't have the promise of making the creators money.

Being able to create manim animations at scale is a value-add, but doesn't seem enough of a value-add to create a business which does anything other than active harm. But it seems to be trying to be one.

blagie··on What did SFFA vs. Harvard reveal about admissions?
40+% of the slots at Harvard are taken up by ALDCs. That's most of the white students slots. If you're like, say, our vice president, you have quite literally a snowball's chance in hell of making it in.

I doubt people are incentivized the way you think they are. Smug attitude aside, everyone wants best for their kids. Many people don't know how. They don't know how because their parents, grandparents, and no one in their community knows how. You end up in what research calls a culture of poverty.

That common among, but not exclusive to, many low-income black communities, due to a history of Jim Crow and slavery, but also in many immigrant communities, and white communities in places like, again, where our vice president comes from.

Improving outcomes for such communities means:

1) Improved economic growth

2) Improved social stability

Representation at Harvard means such communities might trust science. I have very little sympathy for white liberal university faculty complaining how some groups don't believe in climate change when those same faculty are giving those same groups every reason to distrust the academic establishment.

There's also no law which says schools should be meritocratic.

Admissions departments can and do optimize for things like future donations, political influence, and so on.

blagie··on What did SFFA vs. Harvard reveal about admissions?
I'm curious: Which state?

However, there are *many* such effects. Whack one, and I can point to a dozen others.

Until and unless they're systematically addressed, it's actually very hard to either make progress or even draw good conclusions.

What's needed is a project to address *all* of the obvious ones, comprehensively, ideally on a national scale, but initially, at least on a pilot scale.

blagie··on What did SFFA vs. Harvard reveal about admissions?
This is not an accident.

If you talk to admissions officers, over drinks, they'll tell you that poor black students from marginalized US communities have abysmal performance at elite schools. The decision to admit rich black students is very deliberate.

There are many reasons why. Schools in the US are locally-funded, and a marginalized community might have 25% of the per-student spending of a rich community (actual, typical figures). Until inequities like that are addressed, there is little elite education can do to help, even if it wanted, which it doesn't.

The point of an exclusive university is to be, well, maximally exclusive. That's the literal opposite of inclusive.

blagie··on What did SFFA vs. Harvard reveal about admissions?
The open question is the meaning of "proper representation:"

- One which maximizes Harvard's power network?

- Most likely to succeed in the future?

- Representative of the general public by race?

- Representative of the general public in other ways (e.g. religion, political affiliation)?

- Academically, best-qualified right now?

- Academically, most potential? At birth? Now?

- Most likely to donate?

Etc.

However defined, Harvard has much modest discrimination by race compared to discrimination by religion, political affiliation, or home town.

Harvard is 95% liberal, and much more so at the faculty level. If you look at the further-out evangelical, born-again Christian groups, or more conservative Muslim groups, the discrimination is even more extreme. Atheists are vastly over-represented, in contrast.

Until one is aligned on goals and definitions, one can't speak of goals being met or unmet. Harvard can be -- and indeed, is -- racist against blacks by one definition, against whites by another, against Asians by a third, and so on.

blagie··on The intracies of modern camera lens repair (2024)
Only partially true.

In well-designed microelectronics, they will.

The standard circuit involves a fuse, a fast Zener clamp, and sometimes a small resistor (e.g. 1 ohm) and/or capacitor. The design parameter is that, with the current limit from the resistor, the Zener should not blow out before the fuse.

The resistor needs to be small enough to not lose a lot of voltage under normal operations, but to still protect the Zener during the short surge during which the fuse blows. For most microelectronics, that's not hard. A 0.5W USB device might have 100mA of current max, which across 1 ohm is 100mV, so negligible for most purposes.

With high-power devices, it gets more complex.

Of course, consumer devices (a) will never be fixed (b) don't sell on this (c) every penny counts, so there's no market pressure to do things right.

But that's how we used to do it, and how it's still done many places where things count. If I'm building a one-off or few-off, it definitely will have proper protection.

blagie··on Mercurial, 20 years and counting: how are we still alive and kicking? [video]
When I used it a decade ago, virtually everything in mercurial was slightly better-designed, more user-friendly, and more polished. Much shorter learning curve.
blagie··on European Stagnation Is Real
Growing the economy is not "like a bicycle, the slower it goes the less stable it is."

See everything from ancient Egypt to Soviet-style Communism. Periods of time in Ancient Rome would fit in too. The problem with those -- except for a repressive political system -- was that we could not produce enough to satisfy everyone's basic needs.

Today, we can.

Much of the US-style growth is driven precisely by instability. The system feeds on debt. That forces stress. That forces growth. That also means one cannot be "better stewards of one’s self, family, friends."

And so far, no one has explained how "we don’t foolishly bully human nature with policy." Legal and economic frameworks guide what humans do; that's either something we think through and engineer, or it gets engineered for us by market forces, but it's there anyway.

The most freedom-preserving policy isn't a command system, but it's not a pure market system either. And for stability, we can contrast stable economies to boom-and-bust cycles and successful democracies to the very many failed market-economy democracies across e.g. Africa and South America.

blagie··on European Stagnation Is Real
Honestly, the world could use more stagnation.

We've had a lot of change in the past hundred years, and we've reached a point where we can clothe, feed, and house everyone in the world. Once that's done, moving forward more thoughtfully makes a lot of sense.

Who cares if we reach Mars in 2100 or 2200?

On the other hand, I care a lot about avoiding the nuclear / bio / chemical / environmental / and now AI apocalypse.

blagie··on Innocent woman jailed after being misidentified using AI facial recognition
Nope.

https://abovethelaw.com/2016/02/criminally-yours-indicting-a...

You can be arrested, indicted, and held in jail on pretrial, and there is literally no recourse. There are many other ways jail can happen without due process. Where I live:

* Civil contempt. Absolutely immunity. No due process. Record is about 16 years. Having a bad day? Judge can toss you in jail.

* "Dangerous." Half a year. No due process. He-said she-said.

* "Insane." Psychiatric hold. Three days. Due process on paper, not in practice. Police in my town can and do use this if they don't like you.

Absolutely no recourse. You come out with a gap in income, employment, and, if you missed rent/mortgage, no home. Landlords will simply throw your stuff away too.

You're also basically damned if things do move forward, since from jail, you have no access to evidence, to internet (for legal research), and no reasonable way to recruit a lawyer (and, for most people, pay for one).

Can happen to anyone. Less common if you're rich and can afford a good lawyer, but far from uncommon.

blagie··on Iran's internet blackout may become permanent, with access for elites only
> Then Iranians will be reminded how peaceful and prosperous the most other Muslim countries are.

This is factually incorrect. Top 10 majority-Muslim countries, sorted by population:

Indonesia, Pakistan, Egypt, Turkey, Algeria, Sudan, Iraq, Afghanistan, Morocco, Saudi Arabia

Now, the majority of those have problems with seeds in Western Imperialism, but the point is (a) the majority of those have problems (b) Iran's problems also have seeds in US interventions.

The gap between how peaceful and educated most people are, and how bad governments are, is a phenomenon almost unique here. Figuring out how to bridge that gap is the major challenge. The trick would be establishing a collective caliphate -- where the caliph isn't an individual but an institution -- and which spans the Muslim world.

blagie··on Report: Microsoft kills official way to activate Windows 11/10 without internet
Really?

When did this come up?

I know tons of people who run Windows unactivated. The key difference is there's a watermark. Otherwise, it seems to work fine.

blagie··on Meta Segment Anything Model Audio
It does, but my comment was "even in big-O."

Environments might mean the difference between e.g. 16GB and 24GB, but not 16GB and 160GB.

blagie··on Meta Segment Anything Model Audio
For future ML developers: A post like this should include system requirements.

It's not clear from the blog post, the git page, and most other places if this will run on, even in big-O:

* CPU

* 16GB GPU

* 240GB server (of the type most business can afford)

* Meta/Google/Open AI/Anthropic-style data center

blagie··on Adafruit: Arduino’s Rules Are ‘Incompatible With Open Source’
FWIW: I've used Arduinos, but never with their IDE.

AVR was super-developer-friendly well before the Arduino. It replaced the PIC for a lot of hobbyist projects.

To the points in the thread, on major product development, these things don't matter. On the long tail of smaller products, as well as on unexpected successes, they do.

blagie··on Adafruit: Arduino’s Rules Are ‘Incompatible With Open Source’
I, respectfully, disagree with this analysis.

Prototyping platforms have tiny markets, but lead to downstream sales. Many a company were brought down by more developer-friendly platforms ignoring the "tiny" userbase of people who want to do unconventional things.

Most IC vendors provide free samples and support because of this. That's a market size of close to zero -- electronic engineers -- but leads to a market size of "massive." I can get an application engineer to visit my office for free to help me develop if I want.

Arguably, iPhone and Android won by supporting the tiny market of developers, who went on to build an ecosystem of applications, some long-tail, and some unexpected successes.

And arguably, x86 won for the same reason.

Atmel had shipped 500 million AVR flash microcontrollers, due in large part to the ecosystem created by Arduino.

Balmer said "Developers! developers! developers!" Visual Studio was not a major revenue driver for Microsoft; what was developed in it was.

blagie··on Django 6
There is nothing you can't do, given a Turing-complete language.

That doesn't make it reasonable or convenient to do so, though.

blagie··on Django 6
It is, and intentionally so.

NiCd batteries could also sit on a shelf forever, holding their charge. They had virtually no self-discharge, which was super-convenient.

They came in standard form factors (AA, AAA, 9V, etc.).

I really liked NiCd batteries.

But realistically, you couldn't sell a phone or laptop in 2025 which ran on them.

blagie··on Django 6
"Fun" isn't the same thing as "functional."

I remember having great fun in QuickBASIC. And my son enjoys Scratch.

Django code is much more fun to work with than Node, but I can't imagine developing something competitive in it in 2025 to what I'm developing in Node. Node is a pain in the butt, but at the end of the day, competitiveness is about what you deliver to the user, not how much fun you have along the way.

* I think the most fundamental problems are developer-base/libraries and being able to use the same code client-side and server-side.

* Django was also written around the concept of views and templates and similar, rather than client-side web apps, and the structure reflects that.

* While it supports async and web sockets, those aren't as deep in the DNA as for most Node (or even aiohttp) apps.

* Everything I do now is reactive. That's just a better way to work than compiling a page with templates.

I won't even mention mobile. But how you add that is a big difference too.

It's very battery-included, but many of the batteries (e.g. server-side templating language) are 2005-era nickel cadmium rather than 2025-era lithium ion.

I would love to see a modern Node framework as pleasant to work with, thought-out, engineered, documented, supported, designed, etc. as well as Django, but we're nowhere close to there yet.

blagie··on Way past its prime: how did Amazon get so rubbish?
Amazon broke in 2020, when most shopping went online. It never recovered.

I doubt it ever will. Trust takes a long time to earn, and a little bit of time to break. I had four or five incidents on Amazon, cancelled Prime, and I doubt it will ever make business sense for Amazon to get me back.

I do think there's a place for a competitor to Amazon right now which looks more like the old Amazon.

Starting one would be super-capital-intensive. It's not a lean startup. There's only a handful of organizations with the capital to do that, and I doubt any of them will, in fact, do it.

blagie··on Way past its prime: how did Amazon get so rubbish?
It's double-return-fraud.

Amazon shouldn't sell returned products as "new," but as "open box."

The other way it happens is co-mingling. Some vendor sends an "open box" product to Amazon as new, or a fake product, and Amazon ships it out when sold by Amazon since it considers goods to be fungible.

I stopped buying anything which goes in my body from eBay, Amazon, and similar after receiving a premium food product with very clearly fake packaging.

blagie··on The Amazon Kindle War Against Piracy
Bad DRM led to Napster led to Netflix lead to a fragmentation of services led to a resurgence of piracy.

Similar thing happened with music, only rather than piracy, it landed on legal / free (e.g. Youtube). Youtube is just starting to do the consumer-unfriendly thing (but it's got a long ways to go before piracy comes out competitive).

Similar in books.

I'll mention: A lot of these are consumer-unfriendly in some ways (e.g. Netflix DRM), but friendly in others. $20/month for all the movies you can watch beats piracy.

blagie··on The Amazon Kindle War Against Piracy
The cycle has been:

Piracy -> Friendly ways to buy -> Unfriendly ways to buy -> Piracy -> ...

Unfortunately, giving money back to writers involves hopping through piracy. At that point, a new, consumer-friendly service will sprout up. Everyone will use it.

Over time, the service will want to profit-maximize, and will adopt anti-consumer techniques. Leading people to go to Pirate Bay. Leading to friendly services.

Rinse, repeat.

blagie··on Intel Arc Pro B50 GPU Launched at $349 for Compact Workstations
I don't know the full set of differentiators.

For me (not quite at the A1000 level, but just above -- still in the prosumer price range), a major one is ECC.

Thermals and size are a bit better too, but I don't see that as $500 better. I actually don't see (m)any meaningful reasons to step up to an Ax000 series if you don't need ECC, but I'd love to hear otherwise.

blagie··on Intel Arc Pro B50 GPU Launched at $349 for Compact Workstations
Four of these together should, in the abstract, let you run 200GB models, which is where things get very, very interesting.

The biggest Deepseek V2 models would just fit, as would some of the giant Meta open source models. Those have rather pleasant performance.

In theory, how feasible is that?

I feel like the software stack might be like a Jenga tower. And PCIe limitations might hit pretty hard.

← PreviousPage 3 of 34Next →