It's a bit unfortunate that [1] doesn't seem to mention Fleece[2], since Fleece is an already existing, solid format with very similar properties (JSON-compatible data format, uses offsets for sharing, allows for cheap updates, and uses binary search on arrays to achieve O(ln n) lookups in large tables or arrays, etc.)
So many little things for parents. Your list is good. I was also struck by "early baby", for example, for subtext children will only get when they re-watch in 20 years.
So you can't even debug or check the LLM's output for correctness? Good luck with that.
If anything, the best use for LLMs is to produce code in very strong languages like Lean, where correctness of the code can be established. Just trusting machine code output would be irresponsible.
Where do you see sunshine? Are the "audits" public at all? It's indistinguishable from just having Musk kill every agency he dislikes with no further proof, while siphoning data to some unspecified server. The timeline also hints at no serious audit taking place.
Can you point to a country that publicly releases everybody's tax returns, as a policy? I know some countries that release elected officials' returns, but not those of common people.
You think that individual choice is the one and only way to fix climate change? The new regime's policy is "drill baby drill". That's a lot more impactful than people electing not to buy a particular brand of electric cars.
Why does this one man, then, work for a government bent on ignoring science and denying climate change? I'm just laughing at the idea that the grievances are "petty" when the guy is leading a fascist coup right now.
I sincerely hope this is satire (it sure is very HN in nature). "AI" in its current generative incarnation is prone to hallucinations/confabulations that cannot be avoided. In what world is that compatible with a job where a mistake can kill hundreds of people a few minutes or seconds later?
Maybe you're right, that's quite sad. I found dasl.ing to be quite nice and simple as a foundation for dCBOR42, without needing the full ipfs craziness. Oh well.
I think they should have not reinvented the wheel ;-) and use dCBOR42, which has an actual canonical form. But somehow people like to use json in places where it's shown again and again it's a terrible choice.
Can nostr be used for the same form of social network as bluesky (ie, a twitter clone)? It seems that it would only show messages from people the user explicitly follows, for example, and not replies by a third user.
Also it's distasteful to do any sort of content addressing on json data. One would think they'd learn and use CBOR after seeing secure scuttlebutt, but no? Now you have to worry about only sending text payloads, escaping some characters, avoiding whitespace when printing the json... Guaranteed to be a source of bugs...