The justice system can generally deal with gray areas like this. For example the parents of school shooters are usually not held liable for the crimes their kids commit. It depends on a lot of variables.
I've had trouble with the sandbox functionality baked into agents being able to do what I want, particularly Gemini CLI. Being able to write your own .sb file is more powerful and portable.
Claude Code seemed to be able to reach outside its own sandbox sometimes, so I lost trust in it. Manually wrapping it in sandbox-exec solved the issue.
60kmph / 37mph is very fast for somebody who might just be wearing a bicycle helmet (hopefully). If traffic is going that fast, I think it may just not be the appropriate place for a bicycle to be. I've gone that fast on an e-bike before, and it doesn't feel comfortable nor safe.
I agree with having a good helmet, however to be honest my first motorbike ride and car drive at 60kmh were terrifying. Also many people never bicycle even in a 30kmh limites zone because they don’t feel safe.
But I don’t want to downplay speed, as you noted it’s probably the key: most motorbike death are because speed or loose of control without involving any other vehicle. Also small cylinders (< 50cc) are almost absent in the death toll. If suicidal motorbikes with good helmet are allowed, so should be the bicyclists (with good helmet).
Suicidal motorbikes are allowed with license and insurance though. Not saying that's optimal for public safety, but that's a big distinction.
I think that's the logical line between e-bikes an electric motos: at what power or speed do you want to start requiring some kind of licensing or insurance?
Yeah licensing and plates would be interesting. Although an e-bike is lighter than a scooter and will make less damage to the other person, the driver weight is probably signifiant too.
Not sure how that works in the US but in France (and probably Europe?) everyone supposed to get a "civil responsibility insurance" that will cover many thinks including accidents on non-insured (legal) vehicle.
People ride analog pedal bikes all the time in places with road traffic and they impede that traffic when they are going slower - I’ve known more people hurt because someone tried to pass them when they’re departing a traffic light or needing to turn across traffic than from falling down while going “too fast”. It’s frequently more than getting yelled at when multi-ton vehicles intentionally pass by you so close you feel the wind push you away. Being able to go about 35mph puts you at a pace where someone in a car stuck behind you is much more likely to exhibit a little patience.
EBikes are popular and growing like crazy, especially outside the US. There’s somewhere over 30 million in India alone, estimated to double in five years. Their presence is not going away, even in the US, but it takes serious time and desire to get protected bike lanes built. Where I live there’s 6 grocery stores within 3 miles in either direction - and all on the other side of a 4 lane road. You end up riding in the road for part of the trip, and it’s more dangerous from relatively heavy traffic if you’re going 15 instead of 35 for even that short distance.
It is difficult to know whether going faster is overall safer.
In my experience, some fraction of cars will pass a bicycle under any conditions, no matter what speed that bicycle is going (even if keeping up with traffic above the speed limit), no matter how dangerous it might be, no matter if the bicycle has "taken the lane" leaving no room to pass safely -- for some car drivers, it is about getting ahead of the bicycle.
It weirds me out a bit that Claude is able to reach outside the sandbox during a session. According to the docs this is with user consent. I would feed better with a more rigid safety net, which is why I've been explicitly invoking claude with sandbox-exec.
I got very tired of seeing the same video thumbnails over and over.
It seemed like at some point they were pushing into video, of which there were some good ones they put out, but then they stopped. They kept the video links in the articles but since there are only a handful you'll just see the same ones over and over.
I've probably seen the first 3 or 4 seconds of the one with the Dead Space guy about a hundred times now.
You're right, wrote that from memory. It was EBITDA that surpassed anything Twitter previously had before purchasing it.
> Despite a revenue drop from $5 billion in 2021 to roughly $2.7 billion in 2024, the EBITDA margin surged from 13.6% to 46.3% due to drastic cost-cutting measures and restructuring
I used to feel this way, at least about having the TV do zero processing.
Something that recently changed my viewpoint a little bit was that I was noticing that 24-30 fps content was appearing very choppy. I couldn't figure out why it looked like that. It turns out it's because modern OLED TVs can switch frames very cleanly and rapidly, CRTs or older LCDs were not like that, and their relative slowness in switching frames created a smoothing or blending effect.
Now I'm considering turning back on my TVs motion smoothing. I'm just hoping it doesn't do full-blown frame interpolation that makes everything look like a Mexican soap opera.
All you need to fix that is 3:2 pulldown, which all modern TVs should be able to do.
Unfortunately this is another basic feature that tends to be "branded" on TVs. On my Sony Bravia it's split into a combination of features called Cinemotion and Motionflow.
3:2 pulldown (or other telecine patterns) is what was used to go from 24 FPS film to 30 FPS interlaced NTSC video. Your TV or video player needs to undo that (going back to the original 24 FPS) in order to fix a judder ever 5 frames. But that is not going to fix the inherent choppiness of fast camera movements with 24 FPS film and is also not relevant for most modern content because it is no longer limited to NTSC and can instead give you the original 24 FPS directly.
When I'm selling something used, and I look at prices for other listings, I'm competing with them. I want to have a price that is attractive, given the landscape. I'm not messaging all the other sellers and suggesting we all raise prices by 10%, or having an algorithm run by a third party service do that indirectly.
RealPage made it so that landlords were less competitive and more cooperative. Landlords would share proprietary information and then RealPage would help them all set prices collectively. It's just old fashioned price fixing with a SaaS and an algorithm.
So the difference is using pricing info to beat your competitors vs using pricing info to collude with your competitors.
>I'm not messaging all the other sellers and suggesting we all raise prices by 10%,
The way competitors legally message each other to suggest a price increase is via the prices themselves.
E.g. an airline wants to raise the price of a ticket from New York to Los Angeles from $500 to $530 -- and they secretly want the other airlines to follow them and raise their prices too.
1) The airline submits the price increase to the global travel reservation system that all airlines can see. All the other airlines have computers constantly monitoring all the other airlines' ticket prices and can instantly adjust prices in response.
2) The airline that wants the price increase waits to see how the other airlines respond. Either (1) the competitor airlines keeps their lower prices to "take market share" -- or -- (2) they also raise their prices to match which "maintains status quo of market share" but all competitors get to take advantage of charging the higher price
3) If the other airlines don't match the higher price, the airline that "proposed" the higher price then rolls it back to $500. All this can happen within a few hours.
That's the way competitors "collude" to raise prices out in the open. The publicly visible prices are the messaging system. The loophole here is that the changing prices must be visible because the potential passengers buying the tickets need to see them too.
The above scenario has been studied by various papers and the government. The prices simultaneously act as both a "cost to buy" and as a "message to cooperate".
Legal "collusion" via price signals is easier in concentrated industries with few competitors (e.g. airlines). It's harder for fragmented markets or markets with hundreds-to-thousands of competitors. E.g. a barbershop wanting to raise the price of haircuts by $5 isn't going to get the hundred other barbershops to also raise their prices by $5.
>The above scenario has been studied by various papers and the government. The prices simultaneously act as both a "cost to buy" and as a "message to cooperate".
Yea I mean. A simple watch of movie film "A bueaitful Mind" starring John Nash as math genius russel crowe. Crowe equilibrium or whatever it's called. That scene where the nerds were in the bar trying to get the girl. his friends said let the best man win and crowe said - no - only way to win is we collude. and then they won. Now imagine that -- but it's not russel crowe, it's united airlines.
I mean if you look at companies from that crowe equilibruim perspective and treat them as sophisticated and rational.. one would expect most everything to be rigged!
The enterprise sales cycle is often quite long, though, and often includes a lot of hurdles around compliance, legal, etc. It would take a fairly sustained loss of edge before a lot of enterprises would switch once they're hooked into a given platform. It's interesting to me that Sonnet 4.5 still edges Gemini 3 on SWE bench. This seems to bode well for the trajectory that Anthropic is on.
