spiantino's comments

On: Stats – macOS system monitor in your menu bar

[spiantino]

Stats + Hidden Bar + Brave are my first 3 downloads on any new Mac

[clait → spiantino]

Unfortunate that you mentioned Brave invalidating your previous two suggestions!

[apparent → clait]

Different reaction over here. I've never heard of Stats or Hidden Bar, but as a Brave fan I'm going to go check them out.

[mattrighetti → spiantino]

I did not know about Hidden Bar, will try that out! I've been using Bartender for a long time but that is not free.

[ilyagr → mattrighetti]

I haven't heard about Hidden Bar before, but https://github.com/jordanbaird/Ice is quite nice.

On: Boom XB-1 First Supersonic Flight [video]

[spiantino]

seems too small. where do the passengers go?

[josephcsible → spiantino]

The plane they used in this test is a 1/3 scale version of the final product.

On: The Rate of Return on Everything, 1870–2015 (2019)

[spiantino]

Especially when you consider all the tax affordances related to owning real estate vs. equities.

[JumpCrisscross → spiantino]

> all the tax affordances related to owning real estate vs. equities

The only ones I can think of are depreciation (analogous to capital loss harvesting), 1031 exchanges (loosely analogous to step-up basis; this is the biggest difference) and opportunity zones (analogous to QSBS).

If you borrow against your equities, you can deduct the interest paid on that. That mortgage-interest deductions are bigger is a function of the lending being federally guaranteed more than tax law.

On: What Extropic is building

[spiantino]

Obvious grifty nonsense.

On: What Extropic is building

[spiantino]

So exciting! We'd be walking amongst our GAI brethren this very day if it weren't for the computational limits of those pesky RNGs!

[pclmulqdq → spiantino]

I can sell you a solution to that in AWS/Azure (or on prem) today if you really want to use a TRNG for your ML training :)

They are very energy efficient (measured in pJ/bit), but non-cryptographic PRNGs, which are typical for ML, are far more efficient.

It's not obviously wrong to think that AI algorithms will pick up bias from "overfitting" to their PRNGs used during training, but I'm not expecting the benefits to be very large.

On: Gemma: New Open Models

[spiantino]

out of curiosity, why is this a "terms" and not a license? I'm used to reading and understanding the software as coming with a license to use it. Do the terms give us license to use this explicitly?

[turnsout → spiantino]

They do, but unlike a known license, these terms are custom and non-standard. Which means I would guide my commercial clients away from this particular model.

On: Gemma: New Open Models

[spiantino]

Maybe a dumb question, but why is there a Terms instead of a license? That feels a little flimsier as an open source offering

On: Sam Altman is still trying to return as OpenAI CEO

[spiantino]

The EA community includes a lot of AI folks, as well as philanthropists like Dustin.

That doesn't mean this is this kind of conspiracy

[JohnFen → spiantino]

But it does cast a pretty dark shadow over the AI community.

On: Historians have an increasingly strong incentive to tell dramatic stories

[spiantino]

I think a slitting mill is a different thing than the iron foundries described here

On: Reddit’s plan to kill third-party apps sparks widespread protests

[spiantino]

If you click the three dots on a recommendation you can select "show less like this" and you won't see recommendations from that subreddit. Also, there is a setting in user settings to prevent recommendations from subs you don't subscribe to from appearing at all, but I would give the algorithm a chance since there are interesting subs you probably aren't aware of.

Also, I wouldn't call making recommendations in home feed a growth hack, fwiw.