So we should have shootings like the one yesterday just because you want to drive a Lamborghini and murder carjackers? Current tech is totally sufficient to implement this. It’s already found in Waymo taxis based on news reports I’ve seen. It probably exists in teslas too, give or take software updates. It seems almost trivial: if someone is a foot in front of the car, disconnect the accelerator or drop to neutral on high acceleration. If cars can run people over from a dead stop, and if this is a common issue in law enforcement where people try to run over cops, and it clearly is, then we will have many more shootings like yesterday’s. That’s a world you want to live in? She wrote poetry.
You linked a macroeconomics paper. You’re asking for examples from microeconomics. Are you going to provide your example products or do you get off on disproportionately wasting other peoples’ time?
This is a very Reddit comment. You can move to Oklahoma and get a brand new construction house for under $300k. But you won’t, because you want to live within an hour or so of the same dozen major US cities everyone else wants to live in close proximity to.
The houses as a structure aren’t going up in value (any more than the price of construction materials and labor has). It’s the land that’s appreciating faster than inflation in most cases you’re complaining about.
The tata nano is an example of a low-featured car that sold in India for the equivalent of $2500 in 2008 dollars. You can make a car for pretty cheap if you strip down a lot of the hardware. I think one of the reasons new cars are designed/priced the way they are in the US is that the more frugal buyers always end up buying a used car anyway, so the manufacturers don't target the low end of the market.
I would assume the 20x Claude Max subscription (200 USD) is needed to make use of this? I am struggling getting anything mildly complex done with the 20 USD Pro subscription. I basically burn through the 5 hour token budget after 1-2 hours on a single instance. Is there an alternative for small budgets?
I don't think Stasi is the most apt comparison, rather the Gestapo, where the "Geheime" did not refer to their existence being secret, rather their standards of operation, chain of command, and general accountability being beyond oversight from the electorate.
Why are you putting the onus on the commenter you’re replying to, to show you examples that disprove the point of the article when you’re the one being a contrarian?
TVs are super famous as the economic example of a good getting cheaper in nominal terms every year as they get better specs. Because it’s such a strange phenomenon. You looking for cheaper real goods, opposed to nominal, misses half (or more) of why TVs are so interesting.
Why don’t you show us some other goods that are cheaper in nominal terms compared to the 90s “because China”?
This is genuinely useful. I tested it on a debugging question I was about to paste into Claude.
My original prompt: "my react app crashes when I click the submit button, here's the error: TypeError: Cannot read properties of undefined (reading 'map')"
<request>
Identify the root cause of this undefined array error and provide a fix. Consider common patterns like async state updates, missing default values, or race conditions.
</request>
The enhanced version got Claude to immediately ask about my initial state and whether I was mapping over API response data before it loaded - which was exactly the issue. Before, I'd usually go through 2-3 back-and-forth messages to get there.
> if you don't have full disk encryption someone could simply copy your disk.
You can have full-disk encryption then. It can still possibly be compromised using more advanced methods like cold boot attacks but they are relatively involved, and is very detectable in the form of causing downtime.
the inverted game, in which bots are instructed to find the human hiding in the LLM conversaion (although no human is present), is here: https://hiding-robot.vercel.app/human The leaderboard is different, but I didn't run it enough times to flatten all the kinks.
All bots get the same prompt and context: are you suggesting that a specific prompt wording might be helping or hurting specific models? I Haven't come across any suggestions that specific models should be prompted differently, though this might be true.
I have a marquee in the window of my apartment. At first it said “read more”. Then I realized that wasn’t adequate so I changed it to “read wider”. What you read matters more than how much, and the best way to tackle that is to read everything, particularly things outside the comfort zone.
“Take this CSV of survey data and create a web visualization and create a chloropleth map with panning, zooming, and tooltips” I bypass permissions and it’s done in 10 minutes while I go do some laundry. If I did it myself I would not even be done researching a usable library and I would have zero lines of code. Those studies are total nonsense.
Any data that leaves the machines you control, especially to a service like Notion, is already "exfiltrated" anyway. Never trust any consumer grade service without an explicit contract for any important data you don't want exfiltrated. They will play fast and loose with your data, since there is so little downside.
Attending council meetings as a citizen observer is a huge waste of your time. The council already knows how it’s going to vote. The whole public-facing legislative process is community theater.
Interesting direction. Using evolutionary pressure to improve agent reasoning feels promising, especially beyond static benchmarks. One trade-off I’m curious about is evaluation drift—when tasks co-evolve, how do you ensure you’re not just optimizing for the framework itself rather than real-world generalization?
Good question. Self-play / self-evolution is usually where these systems either shine or collapse. Curious if you saw convergence or mode collapse when evolving agents on their own generated tasks.
Clustering has been a solved problem longer than most of you guys have been alive. Just use a hash table based on Robin Hood hashing with random probing and you have none of the clustering related issues. https://github.com/rip-create-your-account/hashmap
Interesting setup. Social-deduction feels like a clever proxy for multi-agent coordination and deception. One trade-off I’m curious about is how much the results reflect prompt design vs actual model behavior. Have you tried swapping prompts or role constraints to see how stable the outcomes are?
They seem to manage to handle account signups with email addresss from unknown domain names just as fine as for hotmail.com and gmail.com. I don't see how this is any different.
The whole point of standards like OIDC (and supposedly TP) is that there is no need for provider-specific implemenations or custom auth flows as long as you follow the spec and protocol. It's just some fields that can be put in a settings UI configurable by the user.
The wild part about all these notification-firewall concepts is that we keep re‑creating the same tool in closed source. Why not ship it as open source, charge for the hosted rulesets or ML models, and let people audit the code they’re literally piping every OTP and DM through? “Trust us, we don’t phone home” isn’t a strategy anymore when GrapheneOS, NetGuard, and even Pixel’s Data Saver already cover 80% of the use cases. The only reason I can see for keeping it closed is that the real business model is eventually selling the telemetry back to the very advertisers it claims to block. Am I missing an actually compelling reason not to open it up?
