sbierwagen
sbierwagen on Hacker News
Human code review? What is this, 2025? The modality today is write with one LLM, review by a different one, (important: two different model families will catch errors one series won't) then deploy right to production.
https://en.wikipedia.org/wiki/Passive_daytime_radiative_cool... (2018)
"If this paint can make any surface cooler than ambient, why isn't it used everywhere?" Condensation is a serious problem, and then also the fact that seven months out of the year in temperate climates you don't want your building's exterior envelope to cool off even faster!
Specifically, oil cooled Porsches take that much, because they have to fill the oil cooler at the front and the lines that lead to the engine at the back.
Amusingly enough, someone from a frontier lab could probably answer this empirically. Years ago Microsoft demonstrated training LLMs on synthetic text-- books rewritten by LLMs to be more concise and more accurate.
https://arxiv.org/abs/2306.11644 It is well known that Anthropic extensively uses synthetic text in training. You could probably get good 50tile, 90tile, 99 etc numbers just from the size of the training materials on Anthropic servers.
A policy paper is intended to change policy. That means appealing to the party currently in power.
>thanks to the use of composite materials
Not always. 777X uses a conventional fuselage but has a 6,000 foot cabin compared to the 8,000 foot cabin on the legacy 777.
Answer: they don't, they just work them into the ground because they can.
Mainland China also has the 996 schedule for office workers purely as a cargo cult ritual, forcing people to sit at a desk at midnight and pantomime doing work.
Stripe obviously records data around friendly fraud, (At minimum they implement Visa Compelling Evidence 3.0
https://support.stripe.com/questions/how-does-stripe-support... ) and since you did not include screenshots of the messages sent by Stripe support I suspect they were saying something carefully noncommittal and legally compliant to get you to go away, which then got spun into an outraged blog post.
> Paliney® 6 is an age-hardenable palladium silver-based alloy ideally suited for demanding low current sliding electrical contact applications
Keep in mind that $275 today is the same as $140 in January 2000. Tech gadgets used to be far more expensive, both in real terms and as a percentage of average income.
Depending on how you convert synapse count to parameters, the brain also has something like a thousand trillion parameters. In that light it's pretty darn surprising that an artificial neural network can produce anything like coherent text.
Are you planning on dying before 2036? That's one estimate for when they'll run out of power.
Fun trivia (well, perhaps not fun) in the second paragraph: "the Long Duration Exposure Facility (LDEF), which was retrieved in 1990 after spending 68 months in LEO"
Long exposure, 68 months, right. But it was only supposed to be in orbit for 11! Challenger being destroyed on reentry made a mess of things.
https://en.wikipedia.org/wiki/Long_Duration_Exposure_Facilit...
>It was placed in low Earth orbit by Space Shuttle Challenger in April 1984. [...] At LDEF's launch, retrieval was scheduled for March 19, 1985, eleven months after deployment.[4] Schedules slipped, postponing the retrieval mission first to 1986, then indefinitely due to the Challenger disaster. After 5.7 years its orbit had decayed to about 175 nautical miles (324 km) and it was likely to burn up on reentry in a little over a month.[6][9]: 15
On a smartphone, you're trusting Apple or Google to make the OS. They already can do anything they want with your system. Do you read every line of code in every security update?
The point of AI is that it's supposed to be intelligent. Why silo it in an app? Instead of telling it what to automate, shouldn't it sit at the OS level, watch everything you do, and figure out what to automate by itself?
From the author's writeup:
>the final pre-trained model came out to about 340 million parameters, and had a final validation bpb of 0.973. The pretraining process took about five hours on-chip, and cost maybe $35. I had my pretrained model, trained in 6496 steps. Things were proceeding swiftly, and cheaply!
GPT-3 had 175,000 million parameters. The smallest of the Gemma 4 models released today clock in at 5,000 million parameters, and I would bet that Google trained them for more than five hours. Just too small and not trained for enough time. A fun art project but not a functional LLM.
Note for the confused: Ecombi achieves this by heating the bricks to dramatically higher temperatures using conventional resistive heating elements, thereby storing more energy, even though the specific heat capacity of any ceramic material is dramatically inferior to that of water.
But, as a result, Ecombi has a much lower system efficiency than a heat pump, since it's essentially just a space heater pointed at a rock. It only makes sense for jurisdictions with time-of-day variable pricing of electricity, and trades off simplicity and low initial purchase price for lifetime cost.
>The drag coefficient was the headline: 12% better than our design target.
Is the drag much better than a regular cubesat? It doesn't look tremendously aerodynamic. From the description I was kind of expecting a design that minimized frontal area.
>Additional surface treatments will improve drag coefficient further.
Is surface drag that much of a contributor at orbital velocity?
From the article:
>Evercoast deployed a 56 camera RGB-D array
Do you know which depth cameras they used?
Ask one of the hundreds of vending machine companies in the NYC area where they put them, I suppose.
https://www.google.com/maps/search/vending+machine/@40.69452...I walked into a Fred Meyer yesterday and saw probably ten vending machines. The Redbox DVD rental machine outside, then capsule toy, Pokemon card and key duplication vending machines, filtered water and lottery ticket machines, Coinstar coin counting machine...
The only people who care about SSO are large enterprises. Coincidentally, large enterprises also are the only customers that make SAAS profitable. Every other plan is part of the sales funnel to the big enterprise contracts.
The author has got to be taking the piss with that page background. Constantly moving blobs, really?
Seems like a foreshock of AGI if the average human is no longer good enough to give feedback directly and the nets instead have to do recursive self improvement themselves.
O365 raising the price to $40 a month ten years from now didn't quite land. Microsoft 365 E5 is $57 a month right now! $100 or $1000 a month makes the joke clearer.