mangecoeur

It would be challenging but theoretically possible. You need very clear skies for the infrared heat not to be trapped by clouds

Even more wild than spectrum slicing is polarisation encoding multiple data streams on the same wavelength.

I find the physics of wavelength-dependent materials properties oddly satisfying. You can do clever things like this, extracting energy from different parts of the spectrum by combining materials and conversion processes. You can beam waste heat back to space through the atmosphere’s “infrared window”. You can explain global warming from first principles. Fascinating stuff!

Llms are far from being ingenious or innovative. The main way they can surprise you today is pulling in information that’s new TO YOU - I impressed with the algorithms they pop out sometimes simply because never formally studied compsci. But when it treads on my home turf I find it very basic, and for bigger picture stuff like system architecture its terrible, by its nature it doesn’t have a “vision” of how the whole system fits together

The first generation were built to produce electricity and plutonium https://en.wikipedia.org/wiki/UNGG_reactor . Civilian-oriented version chosen later.

And manages to ignore that solar power is now the world’s fastest growing source of electricity. Given this is largely driven by economies of scale you could just as well argue if we had gone all in solar in the 70s we would have had today’s solar cost/benefit by the 90s

Something people forget is that France went all in on nuclear also to produce enough weapons grade plutonium for its independent atomic bomb program. Plutonium production was considered a feature rather than a bug of the reactor program.

They absolutely do this.

Lucky for AI companies that OSS projects are doing this, otherwise where will they get their carefully created training data from? They would be reduced to training off the slop their tools produce! /s

Ok i get that some people really do need to deal with extreme temperatures regularly, but it totally misses the point when discussing a mass market car like the Corolla. It’s like asking business casual footware to be equipped with crampons for when you need to walk up a glacier.

It’s absurd that openai managed to spin their cybersecurity failures, where their own security is not good enough to make sure their own product doesn’t go haywire, as a marketing success.

When a company uses its products to hack other companies that’s a crime. If you do it “by accident” that’s gross negligence, and heads roll. But shout “AI” three times and it becomes a marketing opportunity

I’ve had good results with JetBrains AiChat in that really integrates in their IDEs - you can see it using ide tools and i generally found it spends less time thrashing about compared to Claude.

I mean… they still make sublime text. Works great.

There’s an interesting study to be made on whether there was a reduction in insurance claim rejections in the aftermath, and what the morbidity/mortality impact was.

To the surprise of no one who’s doing any teaching these days.

i.e. it's one AI company that's basically guaranteed to never fail since it has a market niche guaranteed by European companies and governments.

I have written software that saves files and saves to standard geopackage formats.

Microsoft is known to have a lot of inconsistencies between its spec and reality. One important part of adoption open standards formats is that when you have real interoperability and competition between consumers of the format, the spec needs to actually work. HTML and the welcome demise of Internet Explorer specific markup through real competition in the browser space is a case in point

This is not true.

In my company I decided we would not depend on Microsoft products. We use kSuite that offers everything MS does, ~90% cheaper. The one real problem we run into is PowerPoint decks that look different or get corrupted when used across different Office suites, including ones that are focused on Microsoft compatibility like OnlyOffice.

Proprietary formats are a clear lock-in tool, because the cost of dealing with incompatibility is higher than the cost of paying the piper. (Although me being stubborn I made claude rewrite the OOXML to behave, but that's an absurd thing to need to do!)

If open formats were the standard and could be reliably exchanged, it would remove a big reason to ever use Office products.

You know it's an India focused device when the top selling point is "Check the cricket results" ;)

Ksuite is literally a google workspace competitor nothing unfocused about that. My company uses them for everything works well.

The idea that science progresses by lone wolf geniuses disrupting the status quo is simply false. It makes a good story for low budget documentaries, but it is basically never true.

Many of the dumb ideas being hyped in this AI bubble make sense viewed through this lens.

Data centres stirring up opposition? Sell a sci-fi vision that you will move them to Space! And reassure your over-extended investors that the data centre buildout rush you’re committing to isn’t going to get bogged down in protests and lawsuits.

The people hyping this stuff are not stupid, just their real goal (make as much money as possible as quickly as possible) has only a vague relationship to what they claim to be doing.

The problem in practice is that quickly one merges into the other. You start with a markdown readme, then you have markdown documentation for a small project. But then one day you need full documentation for your project with cross links, translations, accessibility. With Markdown you end up bolting these things on and each flavor does it a bit differently.

Perhaps some of the blame can be laid with the poor UX of technically superior systems. restructuredtext (apart from the terrible name) built with Spinx can do impressive things but becomes a huge pain to configure. All the XML-based tools like DocBook are very complete but try to get started actually building something - apart from having to author them in XML (which is already a kind of punishment), then you have to figure out XSLT stylesheets, 2000s-era design Java tools for processing them. And just look at the DocBook landing page! AsciiDoc has improved their onboarding recently but does have the issue of feeling like a markdown-ish alternative that's just a bit different for no clear reason.

Sqlite is a great bit of technology but sometimes I read articles like this and think, maybe they should have used postgres. I you don’t specifically need the “one file portability” aspect of sqlite, or its not embedded (in which case you shouldn’t have concurrency issues), Postgres is easy to get running and solves these problems.

Thats the very definition of “anecdata”. (Anecdotes you mistake for representative data)

As European it really struck me how poor the second hand ev market must be in the US. For the “cheap second hand car to do short distances” we have a lot of options really cheap - older gen Renault Zoe and Peugeot e208s, leafs, even the old hyundai Ioniq which is a bit of a cult favourite for its efficiency

The lengths people go to not to make walkable cities (and insulated buildings, fast trains.. you know, all the stuff that actually exists and works)

"The future is..." well it might be until the next fashion change comes about.

This will happen for sure. Designs are fashions that come and go, like jeans that go from skinny to baggy and back again. I don't think anything is absolute, tastes change and people get used to just about anything, and will mainly gripe when anything changes.

Did my phd around that time and did a project “scaling” my work on a spark cluster. Huge pita and no better than my local setup which was an MBP15 with pandas a postgres (actually I wrote+contributed a big chunk of pandas read_sql at that time to make is postgres compatible using sqlalchemy)