Depends on the use. I'm increasingly using binary formats for things like CPU streamtraces, which can grow very quickly into the hundreds of MBs, and not using a text editor for exploring them. Source files tend to be a few KB, tens of KB if they're in dire need of refactoring, hundreds of KB if they're machine generated (and therefore rarely - but occasionally - hand-edited). As such, I don't mind my text editor having a size limitation too much. It does mean that I can't use it for everything, but most of what I edit is source code of one form or another (including LaTeX).

I do this a lot, but I really dislike the fact that I do it. SSH for text editing is an ugly hack to work around the fact that we still can't do file sharing well. I'd much prefer to use sshfs for the editing and only use SSH when I want to build / run code remotely. If Atom came with a nice file browser for remote files, I can imagine changing my workflow.

Definitely important. Vim can only do somewhat-smart indenting. Its APIs don't allow you to distinguish between indenting for semantic blocks and indenting for alignment. I like to format my code so that the reader can adjust the tab size without breaking the formatting, so, for example:

if (someLongCondition() &&
andTheNextBit())

(Slashcode's 'code' tag seems to eat spaces and tabs and expand nbsp, so I've no idea how to actually do the indenting) Both lines would start with one tab (or more, depending on the current indent level), the second line would then have 4 spaces. The s and the a line up whatever tab width you want to view the code with. I generally prefer 4, lots of people prefer 8, and some prefer 2 and so this allows people to set whatever they want.

Vim's integration with clang for autocompletion is also somewhat clunky. OS X has nice autocompletion APIs in the text view, but I don't know how well these are exposed to JavaScript. It's likely to be a lot easier to add nice autocomplete support to Atom than to Vim though.

Ads are only valuable if someone is willing to pay for them, and people are only willing to pay for them if they think someone is going to see them and be influenced by them. If advertisers know that no one watches channel X then it's hard to make them pay for ad time on it, even if they know that a few million people could potentially watch it.

That sounded odd to me too. MSR is really great at making things better, and MS is really good at completely ignoring everything MSR does when it comes to actually shipping products. It's fairly common to see research from MSR show up in open source projects years before MS notices it and incorporates it into a product. Apparently they've been trying to improve this for the last few years, but it's quite difficult to get researchers involved in technology transfer to the rest of the organisation without damaging their ability to do independent blue-sky research. They have had a few successes (F# came from MSR and seems to be gaining popularity), but not a huge number.

How on earth did this get moderated insightful? MSR is not the Microsoft UI group, it is a well respected research organisation. If you actually want to know what they're working on, pick up the proceedings of pretty much any top tier computer science conference and you'll see a couple of papers from them.

There have been a few cases where new BluRay disks have come with new DRM that has broken software players. Sure, you only need to do a software update, but that's annoying if the reason you're using the software player is that you want to watch some films on your laptop while you're travelling and away from the Internet. Apparently the very long load times are still an issue on a lot of players.

Sounds good, but it's a very odd omission for an initial release of a text editor aimed at developers. Maybe they just figure most developers don't write C-family language code...

I find that a bit hard to believe. I've just launched it and not even given it input focus. Atom is using 55MB (33.1MB private), and there are three Atom Helper processes, each consuming 57.9MB, 34.4MB, and 21.4MB (56.2MB, 20.6MB and 10.4MB private) each. So that's a total of over 100MB for a text editor with one window, one tab, and no files open.

Like EMACS, it comes with a reasonable vim mode, although not a very well tested one (o creates two new lines and switches to insert mode, rather than one). The rest of the (non-vim) key bindings are a bit odd (e.g. command-N creates a new tab, not a new window), but it seems useable.

The thing currently that makes it worse than Vim is the lack of libclang integration for autocomplete. I don't know how easy it is to write add-ons that link to a C library (not very, I'd imagine) and without that the autocompletion will suck for [Objective-]C[++].

The only part of this that's correct is the Google part. Dart is StrongTalk with curly braces. The object model, type system, and core functionality are exactly like StrongTalk, the lead developer on both projects is the same, and the VM is based on the StrongTalk VM (open sourced under a BSD license by Sun).

They were bad. The deleted scenes from Episode II actually make it a much better film - basically, anything establishing characters was deleted. The film could have easily lost some of the gratuitous action sequences (e.g. most of the factory-with-flying-R2D2 scene) and kept the plot and character development if length was such a concern.

I honestly can't tell if you're joking.

We're not just in science fiction land, but in movie fiction land where things don't have to make sense, but...

In space, one of the most difficult things to do is dissipate heat. You can radiate, but you can't convect or conduct heat away from you. For a practical demonstration, get a vacuum flask, fill it with boiling water, put it in the freezer and see how long it takes to cool down.

The Death Star has a massive laser-of-doom weapon, which almost certainly produces a huge amount of waste heat. Dumping that heat is likely to be a priority, because you want the planet you're shooting at to be destroyed, not you. Being able to vent coolant in large amounts quickly from the middle of the station is probably very important.

As the summary says, you typically don't have time as a human to make a conscious rational decision about what to hit in a collision. In contrast, an autonomous car can do a lot of processing in a tenth of a second.

Not necessarily. A larger car can have bigger crumple zones. If its crumple zones are twice the size of the small car, then the acceleration that you'll experience in the collision is a lot less and so there's a greater chance everyone will survive (assuming that the relative impact speeds will be the same).

That contradicts your first point. Are you using your car as a weapon to punish the guilty driver, or are you using it as a means of ensuring your survival? It's quite likely that it would be better to swerve into a car travelling in the same direction as you that hasn't made any errors to avoid hitting an oncoming vehicle that is doing something stupid (like being on the wrong side of the road). The relative velocity of the impact will be considerably less.

I wouldn't say no one else does it. From his description, it sounds exactly like how the world looks after taking magic mushrooms. It's no surprise that a brain injury can permanently create effects that a psychoactive substance can create temporarily...

Privacy issues aside, it's also a UI disaster. Previously, I could switch from URL mode to search mode by hitting tab. It became a reflex - create new tab, focus is in URL bar, hit tab, type search term. It took several months to unlearn that bit of muscle memory. And now, rather than a key press that takes a fraction of a second, I have to rely on some flakey NLP code to determine whether I want a search or a URL. I significant amount of the time, it decides that my search term is actually something that wants to be autocompleted to a previous URL that I've visited, so I end up going to a random site. Or it decides that a search term with a dot in it (try searching for command.com) is a domain name, doesn't find it, and then searches a load of similar things and delivers me to a different random page. I've now got into the habit of hitting space at the end of every search, so it now uses exactly the same amount of key strokes for me as the old design in the best case and is less reliable.