Who's paying for the tokens, and since all inputs and outputs are logged "for abuse and safety" what is stopping anthropic from identifying which account is running the daily tests once it gets popular, and A/Bing it into a different slot?
For some reason claude code hates using the read/write/grep tools and defaults to complex sed, awk, grep Bash:* runs which results in always consistently ignoring auto mode due to the classifier not knowing what to do with a 10-line long bash pipe instead of a read + pre/post read and write hooks being ignored and not firing, what is up with that?
I don't understand this either, it makes commands really hard to parse. It seems like 90% of what claude does could be going through some mcp tool that could be easily sandboxed instead of bash gibberish.
If it was human approval ? Read("./file.name") then Bash("docker compose run unit-tests"), sure, I can read that fine !
As I was reading your post I got another approval request that auto-mode did not understand. It was a single Bash call !!23!! lines tall on a retina mac display with six different variables. If anything this is the cause of approval fatigue and auto mode does nothing helpful when this is every call. I also have in my claude.md and global instructions to use tool calls rather than bash grep/cat/find, which are disregarded.
Yeah I agree, I didn't used to feel approval fatigue but lately the huge bash chains are giving it to me. I haven't even tried auto mode. I'm just sure it has some definition of "safe" that does not align with my own.
I'm not sure when claude started preferring huge bash chains but auto mode tends to decline most of them and drop to a permission prompt -- getting rid of the use of auto mode -- because it refuses to handle anything chained with expansions, variables or similar items
I had claude write a hook to deny `find /`. It tested it thoroughly. Much later I learned that the hook denied `find /` but allowed everything else that existing permissions denied, like running whatever random python/bash code it wanted without asking.
From my time at hyperscalers ... realistic for low power traditional DC (lot of regular rackmount CPU server) ... unrealistic for the power and cooling requirements of inference and crypto mining
Many of the votes are not full ban on DC but ban on massive energy dense DC
1) It doesn't reliably offer round-trips made of 2 one-way tickets from 2 different airlines.
this is an extremely bad thing for the baseline normal person and will fuck over normal users hard because they are not competent enough to handle this. most sites do not do this, kayak only does this with piles of warnings to click through and call it a hacker fare
3) I live in an area served by multiple airports. Similar to my first issue, I feel like Google Flights no longer reliably shows me round-trip itineraries that depart from one airport and arrive an another (e.g. depart from Heathrow, arrive back in London at Gatwick). Once I pick a departure airport it
similar to 1) the least common denominator end users cannot reliable handle a "multi city" flight
Both of those are pretty standard for shorthaul flights in Europe - many airlines simply don't sell return tickets, so there's no difference between booking each leg of a round trip as two tickets from the same airline, or as one each from two different ones.
And even sticking with the same airline it's entirely normal to want to see options for all available airports in or near a city. If you book direct through the airline website you'll tend to be offered flights from, say, London (Any) to Belfast (Any) by default.
The popular European services like Skyscanner handle this properly, and it's mostly taken for granted that it will just work. Google Flights used to, so now looks broken by comparison.
I fly very often in the states and it seems to be the opposite situation -- you get little to no passenger rights for a rebooking or delay so you ideally want everything on a single reservation with a single airline so they will want to rebook you if something falls through. On average I do not even get enough of a food voucher to afford a sandwich.
Outside of NYC, DC and SF/LA Ive not been able to think of a city that has more than one commercial high traffic airport. Interesting that, I assume part of it is due to distance-to-next-airport in Europe too?
Man, I loved ITA Software's Matrix tool back in the day. It's a shame Google hasn't kept it updated, as I understand it they've left it running (and even made some UI changes?) but the backend is now rather stale and doesn't share data with Google Flights. For example, it doesn't know about Southwest Airlines even though Google Flights added them a couple years ago, so for my most commonly flown route ITA Matrix returns "Looks like there's no flights satisfying your query!" even though I know there are several nonstops per day.
Maybe the simplicity thing for the typical person is the reason behind these recent changes, it's just a shame as I really valued the previous behavior and there doesn't seem to be any toggle to get it back.
as of at least a few years ago i remember southwest suing anyone that posts their fares without permission, scrapers, (kiwi maybe?) and they don't participate in data feed redistributions
Yeah, that used to be the case. Now Southwest is on Google Flights, Skyscanner, Kayak, Expedia, you name it. For the past few years Southwest has been undoing pretty much everything that it did differently in favor of following typical industry practices.
I wonder if google flights and ita software are legally distinct entities enough that southwest only signed a redistribution agreement with one of them..
There is no accessibility reason ever for covering up the light or recording. I will accept describing objects in FOV etc but nothing should be persistable or sent to a third party that trains on them and saves them without the consent of everyone in view, a company should build non-criminal glasses without these feature sets and have insurance pay for them.
A camera and some rudimentary software should be a pretty good fix for color blindness. A camera and modern AI is a decent mitigation to partial or perhaps even complete blindness (prior art: Be My Eyes). A microphone, STT, and display would help with deafness.
One of the few rare real reasons to owning and if that's the real reason. In that instance the better approach is that the glasses should be held under a permit.
The camera should be disabled until you obtain a permit.
Drones require a license and agreement to fly, these should be too.
Do people not lie? It's not assumption it's more a tally. I meet you at zero. I can't trust words without facts. The former of what you said builds trust.
I would hope the same that I am trusted after proven and not trusted until known.
We all should be able to trust each other but that's not case in todays world.
I am not saying that. I am saying that the glasses should be under permitted use.
You can fly a drone on your own land. You can wear meta glasses on your land. I don't give a shit about that.
What I am saying is that if you want to wear the glasses in public, get a permit otherwise the camera or any geo-tracking features should be disabled.
It should be traceable to who wears these. I don't want to be filmed in public by some creep with a hidden recording device as to the likes of public not wanting hobbyist drones flown over their head.
You may act responsibly with them and that's great, but I doubt those do. Unfortunately that drags you in to the mix and others who use them legitimately too.
Unless they who wear the glasses tell me why they're wearing the glasses I will make association that you're in the same crowd. You stated you don't, so I believe you.
It was interesting to me at my hotel for EDC -- many of the people were downright doomscrolling kalshi, polymarket like it was tiktok but would never go near the casino floor or touch a machine.
When I learned video blackjack is playing the odds in a way that tickled my brain I’ll admit I enjoyed it. I’m sure for some prediction markets do the same. If you can stay level headed there’s probably easy money to be made on prediction markets trading people’s momentum. Staying level headed isn’t easy.
Bummer that tables on the strip in Vegas have such high minimums. Couldn’t even get myself to play one hand on them (but losing $500 in a matter of minutes on Kalshi or Polymarket seemed like nothing)
But if you never review it, how do you know it’s correct, secure, maintainable, performant, etc? You just trust it’s correct because you told the agent to make it correct?
reply