Do you hate reading most science fiction? Because the ratio of dystopic to utopic scifi is pretty unbalanced.
1,348 karma · joined March 23, 2026
Do you hate reading most science fiction? Because the ratio of dystopic to utopic scifi is pretty unbalanced.
The 3080ti is 912.4 GB/s
I'll be sure to tell the economists.
You're just price insensitive. The point of a Pareto frontier is mapping cost and utility. Not everyone can afford to run fable all day.
The end of IPFS at Shipyard: (Many) IPFS Maintainers Winding Down
It's still a a real loss, those are widely used products, but the signal that Protocol Labs is sending about IPFS in defunding this might be even worse for the ecosystem. Cloudflare, Brave, and now Protocol Labs have abandoned it.
Same here. Techies like to say that data centers have always existed, so people must not be objecting to data centers, it must be something else, something nefarious. People don't object to the floors AWS has in some random office building that looks like every other building. Because those were built with planning by corporations usually seeking their own long-term sustainable self interest. If there were tax breaks given, it was usually during periods of economic growth.
Those aren't the kind of datacenters being pumped out and dropped onto communities today in this mad rush to win the new missile race with the new USSR. People are broke, we're in stagflation, none of these datacenters are bringing much of anything of value uniquely to the people who have to deal with their externalities. Even when they bring their own energy to not overload the grid and raise prices, it's usually an onsite literal jet engine powered by natural gas that screams as loud as, well, ever lived by an airport? Now imagine a low-flying plane exactly over your house, 24/7, forever.
I've had a "datacenter" in my apartment or house for decades. Usually it has been a minitower that runs pretty silently in a closet. I got my hands on a used rackserver with 128GB of DDR4 RAM, you know the kind that have fans as loud as a jet engine. Guess who didn't realize exactly how loud those are, even if they're in a closed closet? Or how much heat they generate, especially in the summer? Or exactly how much a sustained 500 watt load costs in California and how that will bump us over our baseline and into extra charges territory?
Yeah, I did to my family exactly what the data centers people are protesting did to those communities. Had I bought a Mac Studio (with even more compute) that sits in a cabinet, my wife would not have vetoed my data center. But I wanted to cut costs and didn't even think about the externalities, just about what I could do what that sweet sweet 128GB of DDR4.
The whole point of LLM/FMs vs good old fashioned ML is generalization to unknown domains, not just unknown tasks. The hunt for "gotchas" is the hunt for "not in your training data".
The Los Angeles Police Department annual budget is over $2 billion. A couple million a year to flock is a drop in the bucket.
> **Google's** up there at 23.4% primarily because GCP/Vertex/whatever they're offering is a billing nightmare and the only way to set a hard cap and sure you won't wake up the next day with a $10k bill is via OR.
OpenAI absolutely lets you set a hard or soft cap. I've set them and they've worked, I think one time I saw a $-0.01 balance. Their docs were probably written by ChatGPT. Later in that article it says:
> A spend limit can monitor spend without enforcement, or you can enforce it as a hard limit so API requests fail after spend reaches the limit.
And they still allow the most certain limit, which Google does not: have $0 on your deposited balance and don't have auto-reload enabled, you get 4XXed instead of going into debt (or at least by at most a couple cents).
Although for fun, go ask Gemini how you can set a cap on GCP spend. It'll try to write you a pub-sub client.
> OpenAI/Anthropic/etc don’t establish the direct customer relationship.
But they also don't have to pay to support and maintain the direct customer relationship and all the billing, support, liability, etc. The credit card merchant fee (a not insubstantial % at scale) is paid by Openrouter (or Stripe now) and the bill Openrouter pays to OpenAI is transferred probably via wire or something with a lot less fees. Meanwhile if you're paying OpenAI via credit card directly, OpenAI is paying Stripe or whoever.
> And OpenRouter can steer traffic away from those labs over to their own model.
The inverse is also true. If you've deposited money on a Openrouter account, you've taken a very big step across their proprietary moat. Would the proprietary companies rather lose them as a customer forever, or be able to entice them back, which can be done with a single change of a JSON parameter?
> The major labs don’t get personalized data about the user to train on, since it’s aggregated.
(Edit: Looked it up, Openrouter does send a pseudonymous but stable hashed userid upstream, stable across different API keys in the same account [0]. That makes de-identification a lot easier.)
First, a lot of the corporate use is getting ZDRed, so there's nothing to track without going back on zero data retention.
De-identification was trivial for normal web traffic before LLMs. Think of all the crap people's harnesses and claws send without de-identifying. Many coding agents can't help themselves from scanning your home directory.
Sure, the type of person to comment on HN probably has good hygene, but the typical vibe coder doesn't even know what they're leaking.
[0] https://openrouter.ai/docs/cookbook/administration/user-trac...
And per the rankings [0] 18.2% of the tokens sold through went to OpenAI [1]. 3.8% for Anthropic though. Google's up there at 23.4% primarily because GCP/Vertex/whatever they're offering is a billing nightmare and the only way to set a hard cap and sure you won't wake up the next day with a $10k bill is via OR.
[0] https://openrouter.ai/rankings#market-share
[1] Although gpt-oss 20b and 120b are indexed as 'openai' because they're the creator even though they don't serve gpt-oss via openrouter. But neither are in the top 20 of models this month, while Luna and Sol are.
My little temperature and humidity sensors that show an LCD 2x the surface area of a watch and broadcast a BLE packet every minute lasts about 12 months on a CR2032 (~200mah). A CR2016 is ~90mah little less than half. And receiving a BLE packet takes way less energy than sending.
Then you have the on-demand stuff that costs a packet or two, like the lost phone finder -- which is what I'm most excited about.
This is the kind of power tripping that easily corrupts people, especially those who don't have much power outside of work. The US national security apparatus is vast and powerful. Many people get giddy at the thought of inflicting punishment on those who "deserve" it. Act rude to a fast food worker, and you get spit in your food. Lots of people cheer the worker who spit in the food of a customer who is merely rude, impatient, or demanding. Or is guilty of being a cop, politician, rich, etc.
I can easily imagine the kind of CS rep who would delight in putting a customer who didn't just go along with things and spoke up for themselves... thats 'rude' and 'disrespectful' to some. Police are the most notorious for this kind of petty "you will respect my authority" but it is in every industry.
HN / hacker culture celebrates this in the 'Bastard Operator from Hell' [0], the sysop who will ruin your work and life with their IT wizardry, no matter if you're an intern or CEO, if they do not feel respected. Or if you interrupt their gaming with your support call.
Some honest feedback: 10 minutes of searching and reading left my ADHD brain confused about pricing and quotas. It doesn't have any numbers for what I'm getting for $4/8/12/mo. And it took me 3 reads of that page to notice that the tiers are not actually the standard tiers, but pay-what-you-can. I first jumped to the prominent table then hunted for the fine print for the quotas, didn't even the read "there is no difference between the subscriptions besides price." Then read it but my mind thought "OK, all features are available to all tiers, but at what quotas?" A normal VC backed company might hide this because they're trying to pull one over on you...
Also, "Sourcehut’s public alpha... What we consider an alpha passes for a production-ready system to some." That's a choice, I get it, but it makes me feel like I'd be jumping from one unreliable platform to another. Maybe link to your uptime? (And the real github uptime side by side :P)
Because another honest thought: unlike GitLab, which pretty much has the same shaped features and affordances that you can one-click import and things mostly work, this would be a pretty breaking change to that would be difficult to revert. I do see full data autonomy is on the roadmap too though.
> The best parts of Github that make it worth using in the first place all come down to the size of the network
Not intrinsically, only as a side-effect of economies of scale. At least for me. The 'social' features make it more compelling and explain its widespread adoption and thus economies of scale, but not what made github worth using for me at least. It Just Worked (TM) to unite all the work we used to do in the good ol' days via repos, patches, tickets, mailing lists, IRC, build servers, and host servers.
I don't need my own VPS running 24/7, I said maybe 100 mins of CI a month, even if I'm pushing and building while others are commenting and PRing multiple times a day, it's going to still take a fraction of that VPS capacity.
Excuse me, I'm literally asking where I can pay to get the specific cloud/web-based features I listed. I'm trying hard to NOT rely on a free tier because I want to pay to have the entitlement to complain when I don't get what I paid for. You know, the way that commerce has existed for millennia. [0]
I don't know where this aggression is coming from. You seem to want me to stay on the free tier and shut up or go all locally-hosted. Great if either of those work for you, buddy.
[0] https://en.wikipedia.org/wiki/Complaint_tablet_to_Ea-n%C4%81...
Can't even view diffs in the web interface!
I'm out! I am willing to pay ~$5-10/mo for a reliable host I can swap as painlessly as possible. Mostly just building little static sites and little PWAs, no more than 500mb a repo and 10gb total. Need a repo with PRs and issues, where people can register free account and contribute or raise issues, GH actions CI, and pages host to deploy static sites. I don't use more than 100 mins of CI a month. Any recommendations?
edit: GitLab free tier has all I need, but I kind of want to pay and have some kind of reliability (not necessarily an SLA, but). But their next tier is $29/mo, which is too much.
The only consistent position on US politics that won't get you downvoted on HN is "both parties are dumb."
Total global war (WWII style), total economic collapse (1929 style), and other kinds of not-even-once-in-a-generation multi-year catastrophes.
Voters getting angry about the price of gas is not one of them.
Nope. This is an internet urban legend. The Netflix prize was only ever for marketing and recruitment. They never intended to deploy whatever winning weights that win because they overfitted on the training and validation data. The dataset they released was not their actual full dataset. It contained zero personal information, just (userid, movieid, rating, date) which is only ever going to get you so far compared to what they do internally.
And it wasn't "a team" that beat the baseline of a ranking algorithm Netflix never deployed, dozens of teams were within fractions of a point from each other. Hence the over fitting strategy.