It is interesting to see the degree to which Claude voice creeps into normal speech. It also leads to some annoying catch-22’s, like rejecting an em-dash where you might have genuinely written one to avoid the appearance of LLM text.
Thankfully I’m lazy and have always used hyphens everywhere so at least I’m safe on that point.
Oh very nice, bliss looks pretty slick. I see musefs as being kind of an unglamorous middle man layer that ultimately just has one job, separate the metadata handling from the backing files.
Since the DB is the source of truth you can absolutely version your metadata. The portable mirror use case is definitely one that people run into, but I see musefs more enabling that than handling it directly. It’d be really easy to handle transcoding and art handling just pointing at a musefs mount or by having a plugin for whatever you’re using handle it.
I definitely agree that continuous automation is the way to go, personally I set it up differently though. Whenever I have a new album in my music folder it gets detected and then beets imports it and museFS autoscans it. That handles 99% of cases silently in the background and the other 1% are the ones I’d want surfaced to make a judgement call on anyway.
Sonnet, GPT-5.2, Gemini Flash, in a set of 21 games, where conclusions are drawn from the LLMs self reported reasoning.
This is like writing a paper about kids in a literal sandbox fighting over ‘territory’.
The models employed don’t indicate the actual extents of machine reasoning even as we currently recognize them. They certainly don’t have the metacognition necessary to accurately understand their own reasoning. As we’ve seen with recent papers on how LLMs do math there’s a complete disconnect between actual and reported mechanism.
Is it really though? Now admittedly I’ve never worked with LexisNexis or Westlaw, but would it really be that difficult to have a tool just check if the citations actually exist?
IIIRC in terms of clients mutt (&co) will actually handle “@“ in the local part correctly.
> But the real reason I do that is just because I just like to sit in anger whenever this breaks the user experience because of programming errors or inconsistencies.
Genuinely delighted by the fact that I’m not alone in that.
This has a serious case of in-housing the edge cases that I don’t think many would want to pay the price for.
The problem with DNS per the haiku isn’t that it’s difficult to understand, or even that running your own DNS server is particularly difficult. It’s that coordinating information and exchange at scale is a tricky problem with a lot of non-obvious edge cases and foot guns.
So trying to reduce complexity by sidestepping DNS really doesn’t do that, it just leaves you holding the bag on all the problems that DNS was quietly solving for you in the background.
My interpretation would be that he feels his wife is incredibly loving in a quantity he isn’t able to match (degree) and in a unique way he’s not able to match (kind). General life experience plus the fact that he wrote that tells me he’s probably wrong and his wife would probably say the same about him, but that’s just speculation.
I’ve had one experience with Claude Code so far that genuinely frustrated me, but it did it to such a degree that it wrapped back around to hilarity. It tried to run a command got an error, realized it needed to cd to a different directory first, and then… didn’t do that.
It tried itself several times going “oh, I didn’t actually cd, let me add that and try again”. I tried correcting it several times “you MUST begin the command with `cd dir &&`”. There were a lot of variations back and forth to try to coax out the correct tool call. Including backing up the conversation and trying from earlier in the context.
It refused. Every time. It simply would not include the cd. Genuinely unhinged behavior.
My initial thought was that that would probably be taking the simulationism too far. My second thought was how funny it would be to swap calendars to get out of a truce early.
“My Lord, you can attack them until March of next year!”
I think that largely depends on how you engage with the internet. To continue the metaphor, the internet has an absurdly high noise floor, but it’s easily filtered noise. So if you do that filtering you’re actually left with a pretty high SNR.
I think this article hand waves and side steps what I see as two notable issues.
The main one is aquifer depletion. The consumptive vs withdrawal argument mostly holds water, but consumptive is a sliding scale. Evaporative cooling for data centers is solidly at the far ‘truly consumptive’ end of that scale because the consumed water will not reenter the local watershed. That’s problematic because aquifers are very slow to refill. So this is genuinely a concern in water stressed areas.
The other is the weak growth model. I suspect we’re only going to see faster and faster growth of data centers in coming years, making the consumption there more exponential than linear. Meanwhile the majority of the other consumptive consumers are strongly tied to demographics and population growth is slowing everywhere. For example agricultural water use in the US has held steady or even declined in recent years.
In fairness, part of that agricultural decline in use is from advancement in technology and methodology and we’ll likely see the same with data centers, but those numbers are unpredictable.
On the whole I agree that the concern over data centers in terms of water (and electricity) usage is overblown to an extent, but I think we do need to pay closer attention to the points that actually matter when looking at the situation.
Accelerando and The Quantum Thief by Hannu Rajaniemi (and that series as a whole) are the best examples of how weird the future is going to get I’ve read.
Other series like The Culture are amazing too, but the aforementioned feel possible in a way that others don’t. For me, I can see the causal chains leading from here to there vividly in a way that you don’t get with a lot of other sci-fi.
That combination of plausible weirdness is unique and I’d highly recommend The Quantum Thief to anyone who enjoyed Accelerando or Stross’ other writing.
Hm, with regards to the difficulty of scaling these systems specifically and really the whole topic at hand more generally I don't know that there is sufficient domain specific data to justify a viewpoint either way. I mean how many examples of scaling a national healthcare system (public, private, or otherwise) to support 300+ million people do we have? Literally three, right? The US, China, and India. Each of which has such significantly different circumstances that comparing them is less apples to oranges than it is apples to giraffes to glaciers. That being the case I feel like it's only reasonable to draw upon non-domain specific data wherein so far as I'm aware and in my experience we see a pattern of scaling existing systems being simpler than originating new ones.
Pretty much this. For further consideration it may be worth looking into Daniel Dennett's writing on qualia (notably that they're ineffable and private).