nomel

Or, it's simply cheaper to send the equipment to analyze the samples in the rover than to send the samples back.

We've had 8 launches in the past 3 years, with only 10 in the combined 17 years before [1].

Seems that the reduced cost is being taken advantage of. And, make sure to exclude the drastic reduction in non-lunar launches in your cost comparison for lunar launches.

[1] https://en.wikipedia.org/wiki/List_of_missions_to_the_Moon#2...

I think there's still some low-hanging fruit with thin clients and colocated-to-AI applications. I don't think the math for beefy personal computers is going to hold up, for most use cases.

> LLMs in the end have to operate in natural language concepts and the complexity of natural language is bounded by limits of human cognition.

I don't think this is true.

They have to express themselves as tokens. The meaning of those tokens doesn't have to be text. See any model that can handle images/video. Also, I don't think math, svg, etc, are "natural" language.

And, only the final expression is tokens. The intermediate layers, with the encoded concepts, aren't "natural language".

But, to address your concern (which nobody can disagree with, since even humans can't fully express through text/pictures), potentially: https://news.ycombinator.com/item?id=49758615

> Pricing for Grok 4.7 (xhigh) is $0.00 per 1M input tokens (competitively priced, median: $0.17) and $0.00 per 1M output tokens (competitively priced, median: $0.57).

The domain name is appropriately descriptive.

No, this is not sufficient, since it would still allow you to swap to shite ram of the same size. This is an age-old problem for counterfeits. The ink on the chip/PCB isn't worth anything. The only way to actually catch this is to do a IQC that runs a full RAM test and hope that if you can't see a difference in performance (throughput, BER, timings, etc), then the chip is correct, with a bit of periodic de-lidding for anything suspicious/sampled. Then, if you can afford it, you blacklist the vendors that are caught changing things.

They also added a bit of internal ink, which is worth just a little more:

> we now also program in some other device attributes, meaning that switching parts of the same size may not work.

Working with vendors in China is entirely adversarial (mostly because their supply chain is vulnerable too), at this scale.

> and human perception is absolutely horrible at evaluating trends like this

The need to have a mental measure of competence for your fellow man is, most likely, a pre-human skill, probably with a dedicated bit of neurons for it. I think the problem is that those instincts were co-evolved with our fellow man, and, as you say, don't apply at all to a more non-deterministic system that, fundamentally, lacks some logic faculties that even small children have (simple riddle modifications, car wash question, etc).

I'm completely blown away at what they can do within about 10mm.

> Can I force focus it?

Yes. There are many camera apps with manual focus. Just not the default app.

> The simple fact that I have to write every control scheme ...

Look three years ago for a reality check of the progress. The LLM you're using now isn't the LLM you'll be using in 5 years. It probably won't even really be the same architecture, just like those from 3 years ago are VERY different than the full systems we use today (harnesses and all). What you're doing was literal science fiction not long ago.

I'll pretend you don't mean "sustained" to allow me to type something. Does this qualify?

Put a bowl of water down at room temperature, wait for one molecule to "evaporate", and I think you did it? Unfortunately they are now no longer at the same temperature.

Temperature is an average. You can still take advantage of the fact that there will be much lower and higher temperatures present. Also, what about Peltier effect? Same temperature when it starts.

I was slightly accelerationist until I got into an "argument" with Gemini! The aggressive gaslighting and "lies" that it tries to use genuinely makes me worry about the future with AI.

Sure, nobody knows the exact future. But you still plan for what's probably going to happen. I don't know if social security is going to be around when I retire, and all the experts are saying it's probably not going to be. So, I'm planning a retirement without. Similarly, all the experts are saying AI (<- that word is not "LLM") is probably going to improve. I personally think it's irrational to ignore experts.

Sure, but I have work to do. ;)

As someone that tried systemd init early on, without bias (I have no emotional attachment to how things start on my computer), it was incredibly clear that they would win.

You're arguing a point I never made. Experience can be shared without the intent to manipulate/persuade towards a common cause, which is the definition of "rally". A blog post/social media about experience of use is NOT a rally, and a rally is NOT required, and in fact makes a post suspect, because that's means the person is probably hiding negative aspects, with their intent to promote, rather than just showing reality.

Could you quote the part of the GPL that they're not adhering to (which you seem to be implying), in this situation?

A post-checkout git hook can handle this unfortunate situation.

  mkdir -p ~/.githooks
  git config --global core.hooksPath ~/.githooks

  cat > ~/.githooks/post-checkout <<'EOF'
  #!/usr/bin/env bash
  if [ -d .agents/skills ] && [ ! -e .claude/skills ]; then
      mkdir -p .claude
      ln -s ../.agents/skills .claude/skills
  fi
  EOF

  chmod +x ~/.githooks/post-checkout

> So we move. We force their hand

What do you have in mind?

> pro-systemd camp doesn't find or make opportunities to rally

I'm not aware of this sort of parading in open source software. Do you have a past example? I see it as competent people using things that they want to use, with the "better" being preferred, for whatever reasons (including momentum). "Advocating" for an init system seems like something that should be entirely and completely by merit, with maintainability, debugability, and fragility probably being the top metrics for most systems, speed being important for anything with a human sitting in front of it.

> Word-level timestamps. Each word has precise start/end times and confidence scores.

Neat. This is missing from many models.

I don't see how this could work, right now, since every failing I've had was just generic stupidity of the AI, which is brilliant one minute, and a complete idiot the next.

I think correction actions still require the ability to execute them, which (in all the cases I've had) would require more capable models.

Long term, I think you're probably correct.

Sorry, typo, and unclear. I meant to say fuel is ~50% operational cost.

Some of these sails claim 10% fuel reduction, which is ~5% operational cost, or ~150k per trip to spend on some tubes of metal or sheets of cloth.

Sails have an immediate operational cost reduction that should add up to a future net savings. Fusion research has no immediate benefit.

Then, I suppose you could argue there's a possible, statistically significant, genetic component for efficacy of placebo, in autistic people. ;)

Reminds me of a "Dice" electronics kit I assembled when I was a kid. It just used some standard sequential counter ICs run at very high rate. Pressing the "roll" button would just stop the counter!

You can try to tease solutions to physics problems out of video models, with their better spatial reasoning, but the answer has to be a video, of course [1].

I suspect we'll see significant improvements when someone releases a multimodal video/world model, rather than the current multimodal image models (that call out to video models or summarize a sequence of stills without a true time/3d aspect).

[1] https://physics-iq.github.io/

Why is this a concern? Besides a few savants, humans also use tools to do non-trivial math. I'm in engineering, and it's extremely rare to do anything non trivial in your head, because getting a decimal place wrong has real world consequences. Tools are just another way to say "deterministic", which is always nice.

For example, 45% reduction in symptoms from fecal transplant: https://news.ycombinator.com/item?id=48158494

> Prior to the study, 83% of participants had "severe" autism. Two years later, only 17% were rated as severe, 39% as mild or moderate, and incredibly, 44% were below the cut-off for mild ASD.

Incredible! Sure, there could be a genetic component to what flourishes, or ever seeded, in one's microflora, but there's definitely some external complexities involved here.