The Hash-anchor edit guy! Sincerely great idea, I used it in my own toy harness to good effect. I just checked this out, never tried it before, and its great! Clearly a well-iterated design with good choices made.
It is so refreshing to see real FOSS and not a grift. Simple openrouter api key, and I'm going.
This is what I'm using from now on. You are doing the best work in this space.
We tried to vote for normal 21st century healthcare and the billionaires spammed race-baiting nonsense and backed an unconstitutional fascist to shut it down.
Man, its hard not to have a reaction to another BaaS. Every "pricing" page really goes to show that engineering took a backseat to rent-seeking.
There is an incentive to take advantage of users ignorance rather than instruct them. There is no incentive to make self-hosting easy or secure or sustainable.
This is literally $600/month for 250GB of storage with no SLA. Cool "value add" bro.
I don't see why tokens/$ would suddenly stop dropping. Maybe this is the first time the cost of compute will plateau, but do have any reason to think so?
There is a strong suspicion, especially of people who are skeptical of AI, that the actual price is being severely subsidized. The sense is that it’s an extreme version of growth before revenue. It is questionable if the true cost of training and inference make any of this worthwhile once Anthropic/OpenAI need to stand on their own and make money.
Imagine you open a cookie shop and you are VC funded, so you charge 5¢ for a cookie to attract people.
- Your real cost is $20/cookie. $15 for the fancy retail packaging and presentation, $5 for baking each cookie.
- You get lots of attention, strong profits and go public.
- VC funding is gone so, now instead of charging 5¢, you now need to charge $25 in order to not be in the red.
One of the reasons people think this is the shenanigans that Anthropic is currently playing, quietly tweaking the behavior of Claude Code and whatnot without really telling people. You can see lots of comments online about Claude Code randomly feeling dumber before Anthropic engineers admit they are messing with it.
Imagine you are on the $200/month Max plan. If the sustainable cost of this is several orders of magnitude higher, would enough current users pay something like $3,000/month for what we currently have?
Sure, yeah, I saw grubhub happen too... but this is compute, not cookies. It gets cheaper.
I don't even get what "skeptical of AI" means. We made AI, many companies reliably teach computers every spoken language. I perform my white collar job with a massive AI multiplier to my productivity.
I'm typing this on a machine comparable to Japan's Earth Simulator, a $350M supercomputer.
That isn't true. In a Codex or Claude Code instance, sure... but those are not the main users of APIs. If you are using LLMs in a service for customers, costs matter.
That is immediate-mode graphics. Fine when you are already power-budgeted for 60 frames each second. UIs typically use retained-mode graphics, with persisting regions.
More than that the questions about space based solar vs land solar for data center calculations seemed hollow as they are easily verifiable. He let Elon get away with this admin does not like Solar as an answer instead of what he is doing to convince them otherwise
The argument is permitting and weather proofing are harder than lifting at certain values of scale for each. We’re not there right now. But if Starship pans out we’re at least damn close, particularly if solar-panel fabrication can be done from out-of-well silicates.
I actually do. The math is more strained than anything present. But a lot of people are rejecting it out of hand without doing anything back of the envelope. Truth is, barring a seismic shift in how we permit data centers on the ground, it takes a within-the-envelope decreases in launch costs to make space-based data centers profitable. Which is then just a cheat code for building a Dyson sphere.
> Didnt startship exploded like 10 times by now?
They all explode all the time. Starship has also been consistently improving its suborbital flight characteristics. I don’t see a good argument for a fundamental design fuckup in the data we have.
> But in 30 months they'll be launchign 1 per hour?
This is nonsense. But within ten years? I think so. At least, we don’t have a good reason to reject that with current data. And that would make the cost equation flip to favoring space-based infrastructure. Which, honestly, is not the answer I expected. (I’ve done aerospace stuff for a while. Most of the back-of-the-envelope math fails. It failed for space-based solar power. It failed for asteroid mining. And it currently fails for space-based data centers. But let launch costs dip a bit, or permitting delays and risks rise a bit, and the equation balances sooner than one would think.)
> Changing permits sounds to me a lot easier than building anything in space
Having done a little bit of both, the latter around data centers, I’ll say they’re different kinds of hard.
> Alright, show me
Fair question. But no, I’m still refining my math and making bets on this. But I’ll start working on an HN comment in a few weeks and try to remember to post it back to this thread.
My basic argument is to try pinning out current datacenter costs, pin out lifted costs, and then work out what cost/kg you need to balance the two. Hint: approval time and interest rates are meaningful variables.
> I’ll start working on an HN comment in a few weeks and try to remember to post it back to this thread
iirc HN threads automatically close, due to inactivity and (/or?) based on time since the original post. I wasn’t able to find a thread with the comments still open from 16 days ago, let alone a “few weeks”, but in good faith I’m assuming that you already know that, and aren’t using that as an out to avoid replying, not that anyone is “owed” a reply by you, or by anyone.
This is all to say, I appreciate the thread as a bystander, and would thus naturally eagerly await your reply if and when it arrives before the closure of individual this post’s comment section.
Prompting LLMs for code simply takes more than a couple of weeks to learn.
It takes time to get an intuition for the kinds of problems they've seen in pre-training, what environments it faced in RL, and what kind of bizarre biases and blindspots it has. Learning to google was hard, learning to use other peoples libraries was hard, and its on par with those skills at least.
If there is a well known design pattern you know, thats a great thing to shout out. Knowing what to add to the context takes time and taste. If you are asking for pieces so large that you can't trust them, ask for smaller pieces and their composition. Its a force multiplier, and your taste for abstractions as a programmer is one of the factors.
In early usenet/forum days, the XY problem described users asking for implementation details of their X solution to Y problem, rather than asking how to solve Y. In llm prompting, people fall into the opposite. They have an X implementation they want to see, and rather than ask for it, they describe the Y problem and expect the LLM to arrive at the same X solution. Just ask for the implementation you want.
Asking bots to ask bots seems to be another skill as well.
Let me clarify, I've been using the latest models for the last two weeks, but I've been using AI for about a year now. I know how to prompt. I don't know why people think it's an amazing skill, it's not much different from writing a good ticket.
Writing a good ticket is not a common skill. IMO it seems deceptively easy but usually requires years of experience to understand what to include and express it in the most concise yet unambiguous terms possible for the intended audience.
