Breck Yunits

Why did AIM die? It was so good.

It could even be a path to LLMs without hallucination. A resurgence for Symbolic AI.

Scroll 1 does not meet that goal, but Scroll 2 will. Looks the same, but under the hood has some really interesting designs related to topological sorting, causality, and logic that I think will be highly useful and relevant for science and all that want to built trustworthy information.

It's a pain to replace a parser engine but I think it will be well worth it.

I've been relatively quiet the past 6 weeks working on Scroll 2.

Scroll 2 has a radically simpler design (figured out a few small compounding improvements), that should make it much easier to explain, use, and allow it to be used for huge million - billion+ line projects.

My vision, which I think now more than ever is possible, has been to "build a language for scientists of all ages".

I've meant that in both the sense of "a language so easy kids can use it" to "a language so compelling Planck and Einstein and Feynman would have used it".

Copyright and science go together like water and fire.

I introduce 2 new terms for medical conditions diagnosable with a microscope:

Hypermito: abnormally high mitochondria levels
Hypomito: abnormally low mitochondria levels

In the process of building image database to define baselines. Collaborators welcome!

People are dying from bad information, and you're doing paywalls.

Why would it ever not be open access? How are your funders okay with your work being put behind a paywall?

Are you recieving any taxpayer dollars?

Do better.

Remember the Pringles ad "Once you pop, you can't stop"?

Pistachios are nature's Pringles.

All big software companies:

1) Unethical "IP" laws possibilize monopoly profits.
2) Monopoly profits possibilize talent monopolies.
3) Talent monopolies possibilize leading products.
4) Leading products possibilize reputation laundering.

Software programs put includes on top.

Academic papers put includes on bottom.

Which way do you think is better?

5 second tip for improving git workflows:

If storage and computation was no issue what would blockchains look like?

Yes I think top 3% income bracket is a lot of money and people shouldn't talk about LE as a "non-profit". The avg non-profit salary is $39K.

GitHub - breck7/ScrollHub: The fastest way to publish. The fastest way to publish. Contribute to breck7/ScrollHub development by creating an account on GitHub.

I use it on ScrollHub, which is open source and public domain and you can run on your own server (and use all the LE code I write). LE requires so much more engineering effort than it needs.

I'm not asking for much. 365 days. A 1 byte change lots of people ask for

github.com/breck7/Scrol...

Again, they can change 1 byte. But they won't. I could not understand that. Their technical argument for it is weak. Then I see they're banking a quarter mil a year writing bash scripts and I get it--they want 4x the traffic.

I can certainly keep my mouth shut and not complain about a retarded engineering decision, but that only hurts all the other users who have been complaining about this and who automation is more of a pain.

And then also, you hit LE rate limits very quick and they lecture you even more.

You have never experienced LE rate limits

They could 4x capacity by simply allowing 365 days, which most people want for most use cases.

Again, if certs are being compromised so frequently that you can only trust them for 90 days, we have a much bigger problem.

I'm not against 90 day certs. If you want 90 days, sure! Just give the people what they are clearly asking for, a 365 day option (since they wrote a whole blog post saying people often ask them for 365 days).

Wait so they only provide services to those who give it money? Why aren't they paying taxes then, because that sounds like a for-profit.

Tell me you haven't experienced LetsEncrypts crazy low rate limits without telling me.

Imagine defending this "non-profit"

Setting up Lets Encrypt for 1 site: very easy.

For 1,000 sites: much more complicated, for no technical reason. All because LetsEncrypt wants to write a blog post rather than change 1 byte.

I haven't looked, but I would guess their Director makes a heft "non-profit" salary.

If you are Gmail or Bank of America, sure, have a 90 day cert. Or even a 6 day cert!

My local coffee shop does not need a 90 day cert. If LetsEncrypt Certs are compromised so often that 1 year cert is too long then we've got big problems.

No, it's premature optimization. I challenge anyone at the LetsEncrypt org to a real time discussion of this.

It makes no fucking sense. I guarantee you when we examine their numbers we would see it has zero impact on security and is just about them making themselves more needed.

I'm not against automation. I'm against 90 day certs forcing automation way before builders even know whether the site will be around in a year. >50% of sites don't last a year. People launch 10 ideas for every 1 they stick with. Thus 90% of automation work is WASTED. LE could fix with 1 byte.

99% of websites should be a git repo.