hex4def6

That $1000/kg figure is reprocessing something like 10kg of material to get 1kg of enriched material.

Current costs to launch a kg to orbit are something like $2,000-$6,000.

If we're comparing it to you enriched amount, you'd have to launch 10kg, which means you'd have to hit $100/kg launch costs to break even. I'm not convinced that will ever be possible; that's on the order of a first class arline ticket for a 50kg person.

This also doesn't consider what you're getting for launching it into orbit. If the 100 year risk of a casket leaking is (say) x mrads into the environment, you're going to have to consider what the equivalent of vaporizing all of it into the atmosphere in the event of a failed launch is. the fallout (hah) from a failed launch of nuclear waste seems magnitudes more catastrophic than having the same stuff slowly leaking in the middle of a desert cave.

On the other hand, I used to have an automatic open garbage can. Absolutely hated it. Ended up replacing with a foot operated one. The benefits were:

* Never ran out of batteries

* Didn't rely on an IR sensor that would work about 80% of the time, but would sometimes require waving more than once to get it to open.

* Didn't have a delicate gearing mechanism (automatic one eventually stripped a gear and broke).

* No set open time. God forbid you throw something in, turn to the counter and throw a second thing in...

* large lid, since it relies on foot power, not a motor / battery can only realistically open a certain weight lid.

My inlaws also have an automatic one -- last time I visited I saw that they also suffered from the stripped gear issue...

I wish public restrooms used foot-operated faucets instead of the existing "wave frantically for 3 seconds of water" sensors they have.

I am jealous of the auto-seal garbage bag option, but I am skeptical about long term reliability.

Ah, I think you're correct. Although the watts per square meter still don't make sense.

Many questions.

A couple:

* 1m² = 1x2m? 4m² = 2x4m? etc. I'm confused by this.

* Why 25 m² (5x10 m) = 20.8kW, but 100 m² (10x20 m) = 62.4kW? 4x the size in m², but only ~3x the power? Shouldn't it be 83.2kW, not 62.4kW? It doesn't make sense efficiency would drop....

* You're speccing this as having a 25 year life (with $50/yr maintenance). I'm feeling a bit doubtful that zip-tied tarps under UV and dynamic load are going to last that long, to be honest. The tracking system also seems extremely susceptible to dirt etc.

* This thing looks pretty janky. I'm also not convinced a good 20-year storm wouldn't completely wreck it. Proper wind turbines have the ability to weather-vane into the wind, lock the rotor, etc etc. Again, in it's current incarnation, I'm doubtful it would survive.

* Your "guaranteed Cp of .32" seems... optimistic. Given it looks like you've built actual units, what are your real-world results vs the CFD numbers?

Yep -- I've gotten multiple postcards in the mail and/or product box after ordering something telling me I'd get ~$5-$10 if I left a 5 star review. They thoughtfully included exact screenshots of how to submit the review.

I also went looking for a way to report it. Eventually opened up a chat, didn't really get anywhere.

(I also didn't appreciate getting junk mail for ordering something).

I've been playing around, got it to work, although quality was a bit crappy. Still playing around with the settings that get exposed, but you're welcome to look at : https://huggingface.co/spaces/jonatei/MoebiusDemo

Note that I'm actively messing with it, so it may break for short periods of time :)

It's also running on the free CPU, so it's like 80 seconds per image...

In VR, it's well known that your peripheral vision is a cue to your brain to infer motion direction.

You can give yourself serious motion sickness in VR if you grab hold of the VR camera and force a roll / pitch in this state.

One of the things some games do is during motion create "tunnel vision" when you move, shrinking down the size of the video feed and putting a black border around it. This significantly reduces this unpleasant effect. I assume the high resolution portion of your eyeball isn't used for motion inference, which makes sense.

I'm thinking this feature is exploiting this same peripheral vision cue.

Yep.

Most people aren't in a quiet environment when they listen to music these days. Compression helps significantly with this.

What would be neat would be to have a compression metadata 'guide' that would allow a compressor on-device to perform the compression, rather than baked into the audio track.

This would allow the user to tune 'severity' of compression. In a car / fancy headphones, you could sample the ambient noise level and adjust accordingly.

While that's pithy, I think it's also incorrect, because it implies that Oracle / Ellison is controllable by us, in the same way a tool / lawnmower is. That's absolutely not true. It has its own motivations that are best-case neutral to our goals.

It might be better to think of ourselves as individual fish in a school of fishes, and Oracle is the boat with a mile long dragnet. It doesn't care about the individual fish; it's not worth it's time to consider us individually. It's thinking in terms of tonnage.

I think if it conflicts with a CMD command it's not shipped, but if it conflicts with a powershell command it's ok.

Of course. LLMS still have huge weaknesses in distinguishing between incoming unsanitized data, and their operating instructions.

It's still malware though. Unlike some backdoor that you could plausibly claim was just a simple memory leak, the instructions for this one are literally written in plain english. Wouldn't be very difficult to show intent to a jury with that one...

https://www.flickr.com/photos/blairmc99/55275010559/in/photo...

I don't think I've ever seen a manual that expected me to use DeMorgan's Theorem as part of a test procedure... :)

One of the all time greats. I think I'll play through it this evening.

"...And what is the 'Self', if not a pattern of data? What is consciousness, if not an illusion of intelligence residing within meat?" — Prime Function Aki Zeta-5, "The Fallacies of Self-Awareness"

Agreed.

For such low frequency stuff, it feels way safer to just buy a cheap <$500 scope for this kind of work. Using a $50k scope when it's not needed just seems needlessly risky.

Also, float the DUT, not the scope... Sometimes that's not possible, and the temptation is there, but it's really not worth it. Just buy the right gear like a diff probe. You can get one for a few hundred bucks if you don't mind going downmarket.

You can also use two probes and do CH2 - CH1. (Disconnect the GND clips!)

Hmmm.

Looking at: https://journals.plos.org/plosone/article/figure?id=10.1371/...

I'm not sure I believe the graphs.

For example, here's another frequency response chart of some stethoscopes: https://www.researchgate.net/figure/a-Frequency-response-of-...

How is it that professional stethoscopes can be that different, and yet this 3D printed one can match a gold-standard one almost exactly?

From what I can tell there's no audio engineering / modelling that's been done here -- It's just some crude openSCAD tubes. And it's not even optimized for 3D printing; a 3D printed tube with a circular cross-section is going to have bridging issues at the top which will result in internal roughness. I have to imagine that results in attenuation. (A better internal shape for a tube is something that looks like "ô". The ^ will print much better)

The type of plastic used and its frequency response, the thickness / stiffness of the silicone tubing, the height / width of the bell... There are so many variables that I think would make significant differences in performance. The fact that they see basically no difference is highly suspect.

This feels like one of those "3D-print everything" fads that was popular a few years ago. Yes, you can make a 3D-printed adjustable wrench, but even the most miserable dollar-tree metal version will beat it in every possible metric.

Likewise, on Alibaba, if you order 200 pieces, I'm seeing metal ones as low as $1.22/pc. I don't believe that this 3D printed one will even be as good as those.

I'm curious about the units -- watts doesn't really seem right.

It seems like joules would make more sense, right?

I think you could alter the prompt in subtle ways; a period goes to an ellipses, extra commas, synonyms, occasional double-spaces, etc.

Enough that the prompt is different at a token-level, but not enough that the meaning changes.

It would be very difficult for them to catch that, especially if the prompts were not made public.

Run the variations enough times per day, and you'd get some statistical significance.

The guess the fuzzy part is judging the output.

I was in the same boat.

I'm in the bay area, did a roof mount system. ~9.5kW with Enphase microinverters.

I think it cost be ~15k all in. The usual installer price is normally around $2.5-$3/w (so about $25k - $30k).

I got someone to do the plans (because I was under the gun to get on NEM2), but that was $300 well spent.

Did it all myself -- mounting, running conduit / wiring etc. Wasn't too bad. Probably about 3-4 Saturdays of work.

Here's my install log: https://www.reddit.com/r/diySolar/comments/1c6jfjv/finally_d...

The point of a benchmark is that it allows a relative comparison. The Pelican is one such benchmark.

Feel free to create a "how does it compare to Claude 3.5 Sonnet" benchmark. If people find it useful, it will be run against new LLMs to generate additional points of comparison.

I will also say; it's really easy to just skim past comments. I suspect your ROI time-wise in creating this account to complain will never be recouped compared with just skimming past pelican comment chains.

"The memo stated Gov. Newsom's administration made two errors. The first involved double counting CalPERS contribution rates for the upcoming year, which the LAO said was a $1.6 billion miscalculation. The second issue involved incorrect contribution rates when the administration calculated how much money the state would need to contribute to CalPERS in the years ahead. The LAO stated that mistake amounts to about $450 million. "

...

"This isn’t a calculation error – it’s revision to better estimate how these payments are made," said H.D. Palmer, a spokesman for Newsom's Department of Finance. "We told legislative leaders and the LAO back in February that we would update how we estimate these payments once this issue was identified. We’ve already made that adjustment, and it will be reflected in the revised budget next month."

Can someone please explain to me how double-counting isn't a calculation error? Best attempt wins.

When a political organization has no qualms about putting out a statement like that, it's a sign that they do not respect you.

First of all: Awesome work! Playing with it now.

One suggestion: The main splash screen image is nearly 8MB big. It takes a noticeable time to download on my connection. I'm not sure what bandwidth costs these days, but seems like that could be something to optimize.

Sure.

But I bet you'd have a significantly easier time converting a child rather than a 30/40/50-yr old to a religion.

My point is that LLMs are suggestible, perhaps more so than the average adult, but less so than I child I suspect. I don't think suggestibility really solves the problem of whether something has AGI or not. To me, on the contrary, it seems like to be intelligent and adaptable you need to be able to modify your world model. How easily you are fooled is a function of how mature / data-rich your existing world model is.

That just implies LLMs are suggestible. The same is true of children. As we get older and build a more complete world model in our heads, it's harder to get us to believe things which go against that model.

Tell a 5-yr old about Santa, and they will believe it sincerely. Do the same with a 30-year old immigrant who has never heard of Santa, and I suspect you'll have a harder time.

That's not because the 5-year old is dumber, but just because their life-experience ("training data") is much more limited.

Even so, trying to convince a modern LLM of something ridiculous is getting harder. I invite you to try telling ChatGPT or Gemini that the president died a week ago and was replaced by a body-double facsimile until January 2027, so that Vance can have a full term. I suspect you'll have significant difficulty.

I've gotta say, this writeup gives me an itchy feeling. It really does feel like poking around a synthetic brain at this point.

You could make the argument it's closer to the blocks of a CPU compared with a brain, and it's no different to copy-pasting some IP block for eg, HW JPEG decoding. But I feel like the difference here is we're 'discovering' these blocks / organs. They weren't designed, they were evolved.

Linux can fool you into that sense of security for a long time. But there will come a point where the facade crashes down.

Maybe it's plugging your laptop into an external projector, or getting to sleep and wake correctly without the WiFi driver segfaulting, or maybe it's trying to get HDR working, or audio routing or...

The author explicitly shows how in Apple's walled garden, even that had hurdles that have been put up, 'for security'.

If you're benchmarking something, old & well-characterized / understood often beats new & un-characterized.

Sure, there may be shortcomings, but they're well understood. The closer you get to the cutting edge, the less characterization data you get to rely on. You need to be able to trust & understand your measurement tool for the results to be meaningful.

Aren't human coders also nondeterministic?

Assigning different agents to have different focuses has worked for me. Especially when you task a code reviewer agent with the goal of critically examining the code. The results will normally be much better than asking the coder agent who will assure you it's "fully tested and production ready"

As someone who worked on kindle power consumption many years ago: One of the (by far) biggest consumers of power is the WiFi connection. It has to wake up and respond to the AP in order to not get disconnected every x seconds.

Off the top of my head, I think 'on' average power consumption was ~700uA without wifi, and about 1.5mA+/- with Wifi. This is from over a decade ago, so my memory is fuzzy though...

Obviously, page changes used relatively large amounts of power. I don't recall the exact amounts, but it was 100s of mA for seconds.

There is also an "every x pages, do a full screen refresh (black to white)" to fix up the ghosting issue that the article writer saw.

Same.

I've been exploring some computer vision recognition stuff. Being able to reason through my ideas with an LLM, and make visualizations like t-SNE to show how far apart a coke can and a bag of cheetos are in feature-space has been mind blowing. ("How much of a difference does tint make for recognition? Implement a slider that can show that can regenerate the 512-D features array and replot the chart")

It's helping me get an intuitive understanding 10x faster than I could reading a textbook.