adinisom

I think focusing on "user facing impact" is a good idea and something I'll be using in the future. One challenge is that developers don't always know the user facing impact of their improvements.

For example if I notice a logic error in a system and fix it, I may not know what if any software configuration could trigger the error. If it were obvious, QA would've found it. And given the choice between further investigation or leaving the bug unfixed, my employer would prefer the latter.

Or one time I was working on a new feature that required speeding up the system's implementation of malloc(). That improves the performance of the entire system to varying degrees. Pinpointing exactly where the user's experience improves would require extensive benchmarking outside the scope of the work.

I marvel at the "mini toolbar" in Microsoft Office. It shows up when selecting text and neatly manages to get out of the way when not needed:

1) Initial toolbar placement does not obscure selected text.

2) Toolbar fades away if mouse does not move towards the toolbar.

I think the idea is that false positives (cert revoked) result in a call home to Apple for an actual check.

The other idea is that because Apple knows all existing certificates, they could conceivably construct filters that have no false positives for those existing certificates... sort of like construction of a perfect hash.

Can confirm... a phone running Google's Live Transcribe propped up next to my laptop is even better for me than Google Meets captions because you can see the history. Will have to try Google Recorder.

Comingling means the buyer can't vet the seller so it's entirely on Amazon to ensure the goods aren't counterfeit.

As you say, Amazon was able to identify the merchant specifically but wasn't able to determine whether the goods are genuine. If the goods weren't comingled, this would be less Amazon's problem and the goods wouldn't be in Amazon's warehouse so at least the seller would still have his inventory if Amazon blocks his products.

Censorship in the form of moderation often makes sense for businesses seeking to maintain their properties. If I run a forum for car enthusiasts, I should be able to curate the community and limit discussions on locomotives.

But what of the platforms that make my forum possible? The Cloudflares the protect from DDoS, the DigitalOceans that provide the servers, the ISPs that provide transit, the Facebooks that provide social media support? Should they have the freedom to decide they don't like my forum and kick me off? I've come to believe that the more general purpose a service is and the more communication it controls, the more harmful the business's freedom to censor is for a free society.

Trademark infringement seems to be a strict liability tort: https://www.tilleke.com/resources/application-strict-liabili...

Could use Geogebra.

https://www.geogebra.org/

Another option:

Fairmount, an app DVD ripping on MacOS, uses WebDAV: https://github.com/BoxOfSnoo/Fairmount

This seems a little more informative:

"AngelCAD is a powerful open source 3D solid modeller based on the Constructive Solid Geometry (CSG) modelling technique, expressed in the AngelScript language. The software creates 3D models in STL or other file formats." (https://arnholm.github.io/angelcad-docs/)

The first is an issue: can only be looking at one thing at a time.

The second doesn't seem to come up much in practice. Can eat and sign at the same time unlike voice.

Third isn't true -- signing is more expressive because it takes advantage of space, facial expressions, and signs can be modulated to adjust the meaning. But it almost has to be since the symbol rate is lower than speech.

FYI, the device above the keyboard is a refreshable braille display. The self-contained computer was probably a braille notetaker... the keys are chorded and six of them represent the six dots in a standard braille cell.

The one in mine signals collision warnings on metal signs above the road. Although in fairness it's never triggered emergency braking.

Removing content when you don't want but are required to do so seems like non-free expression. Although it might be difficult to distinguish between content being removed due to legal requirements vs just not wanting to host the content.

On top of that, people assume estimates are normally distributed. In my experience there's much more skew -- surprises are rarely in your favor.

This means any given "good" estimate is either unlikely since it's in the tail of the distribution or overly optimistic if in the bulk.

> We don't know what a "low viral load" means for infectiousness.

One way the phrase is used refers to severity being dependent on initial viral load. That's what I think you're talking about. Someone can have a mild infection or asymptomatic but still be contagious.

What jsharf is referring to is that below the infectious dose, exposure to the virus doesn't necessarily develop into an infection. This idea is explored here: https://erinbromage.wixsite.com/covid19/post/the-risks-know-...

"possibly" here means that while Y is hosting derived work, even the uploader doesn't know that for a fact since the software doing the uploading doesn't expose this detail.

Knowledge of how to combine them being a derived work is an interesting theory. One challenge is that facts aren't typically copyrightable. On the other hand you could argue contributory infringement instead.

This seems similar to SLIPS: https://wyss.harvard.edu/technology/slips-slippery-liquid-in...

How do LESS and SLIPS compare?

Proper situational awareness while driving is a visual task. As people age, they begin to lose hearing and ability to localize sounds due to age and noise exposure. Leaning on hearing for safety becomes a liability.

From prior experience working with non-UI statecharts:

1) State transitions are essentially "goto". Structured programming offers abstractions that are easier to think about and less fragile to requirement change. But in some situations small state machines can be a win.

2) Statecharts offer "structure" in the form of substates and superstates which are much like inheritance in OOP. In my experience these are fragile to requirement change.

3) Statecharts are advertised as being explicit and therefore easier to check for correctness. They can be, although if there are multiple state machines operating concurrently or even an "outside world" acting on them, the set of possible states is no longer explicit.

That might work, although I think there are two limitations:

1) Hearing aids have a 10ms latency budget. So no matter how much processing they can do, they're limited by how many samples they can look ahead and that limits the design of the filters. The brain can presumably look ahead further to separate sound streams so I think it's pretty impressive that ideal binary masking works.

2) Hearing aids have a power budget. The ones I've looked at achieve low power by running a FIR filter in hardware to shape the sound while a DSP classifies the sound and adjust the filter taps. The DSP doesn't have to run at the same rate as the filter. That seems well matched to the binary filter approach. Likewise features extraction might not run at the same rate as the DNN.