Yes, but don't discount that value. I used to run llama.cpp on the command line because it's amazing that you can ./a-brain. These days I run ollama on the GPU computer and connect to it from my laptop or phone using Zed or Enchanted.
We also expect costs to go down as a technology matures, yet over the first 30ish years the costs of generating a watt of nuclear energy went up two and a half times in constant dollars. A cynic might wonder whether other energy industries may have lobbied for onerous regulations in order to limit the impact of this cheap new source.
All of these guys are woefully small-fry compared to Trump. Their real crimes were pretending that they had the moral high ground, and getting caught.
"People look at a tree and think it comes out of the ground, that plants grow out of the ground, " he says, but "if you ask, where does the substance [of the tree] come from? You find out ... trees come out of the air!"
I welcome being disproven, but I don't think you've done that. Blockchains have a head, and the head is advanced somehow. How does your identity chain advance? Who has the authority to advance it? Without proof of work, you have a Merkle tree. Without rewards, what is the incentive to compete proof of work?
Regarding identity and authorization, you haven't even made an argument. Mine is https://github.com/nostr-protocol/nips/pull/1450
did:plc is a shitcoin. Let me explain to you why.
On #Nostr, your public key is your user ID. On #Bluesky, although you have a public key, your real user ID is a did:plc.
To generate a did:plc, you must make a POST request to https://plc.directory/
Yes, https://plc.directory/ (the literal URL) is part of the protocol definition. They'll say it's not, but it is.
They'll say it's not, because theoretically you could run multiple plc.directory servers, and people could point at different ones. But there are big challenges around security and UX of that, and let's be real, Bluesky never intends to "fix" it (it's not broken).
Now, if we just put the whole thing on a blockchain that would totally solve that problem. But in fact, did:plc "does not make use of any blockchain". Because the accepted decentralized way to achieve this _would_ be to put it on a blockchain, so they had to make that qualifying statement against it.
So, they needed to use a blockchain, but they don't want to use a blockchain. So they invented their own blockchain: a regular webserver!
And they pretended their regular webserver was a blockchain. And this, is why did:plc is a shitcoin.
> Now, if we just put the whole thing on a blockchain that would totally solve that problem.
I would be very surprised if this were true, as I have failed to find a sound use of a blockchain outside of a fungible currency.
The problems are always deciding whether a change is valid, and enforcing the on-chain information. With Bitcoin a change is valid if it sends value less than what's in a wallet to somewhere and is signed by the corresponding private key. The change is also respected because now the wallet doesn't have control of the value, and (probably) someone else does. Both of these are automated and verifiable by anyone.
When you put something else on the chain, say the deed to a house, miners can't validate the change unless it's an ordinal. Fair enough, we can assign an ordinal. But they also can't enforce the effect: only the government decides who owns a house. Once you identify this part which isn't automated, you might as well replace the blockchain with a database.
So, on the one hand, it seems that they did the right thing and avoided using a blockchain in a way that doesn't make sense. But on the other, it's presumably because only one group can determine whether a change is valid, which means it was never distributed in the first place.
Identity is a surprisingly difficult problem. Nostr gets most of it right, except for combining identity and authorization into a single key. No reasonable system still uses "$username:$password" as their session id.
I think Ellul was referring to modern propaganda techniques, given simper ones work quite well. This quote from his Wikipedia page is :chefkiss:
> Also within Propaganda Ellul claims that "it is a fact that excessive data do not enlighten the reader or the listener; they drown him. He cannot remember them all, or coordinate them, or understand them; if he does not want to risk losing his mind, he will merely draw a general picture from them. And the more facts supplied, the more simplistic the image" ... Additionally, people become "caught in a web of facts they have been given. They cannot even form a choice or a judgment in other areas or on other subjects. Thus the mechanisms of modern information induce a sort of hypnosis in the individual, who cannot get out of the field that has been laid out for him by the information" ... "It is not true that he can choose freely with regard to what is presented to him as the truth. And because rational propaganda thus creates an irrational situation, it remains, above all, propaganda—that is, an inner control over the individual by a social force, which means that it deprives him of himself".
The Pope disagrees
This falls under one of my favorite categories:
- Lies
- Damn Lies
- Statistics
So let's say that there are 90 white folk and ten black. When a white person randomly attacks someone who happens to be white... they don't show up on this graph at all. Same for black/black.
Now, ten people randomly attack someone:
- 8 white people attack other white people (not reported)
- 1 white person attacks a black person (10% chance)
- 1 black person attacks a white person (90% chance)
What gets graphed?
- zero white people attacking white people (8 not reported)
- zero black people attacking black people (none occurred)
- "1x" white people attacking black people (this is the reference scale, representing an unscaled value of 1/90 = 0.011)
- "9x" black people attacking white people (an unscaled rate of 1/10= 0.1)
And 9x sounds pretty bad! But it's just a well spun lie. All of these acts were committed at random. What you actually want to graph is how disproportionally likely one group is to attack another group, per capita. This would be 9 attacks on whites, 8 from other whites, and 1 from blacks:
- white: 8/90 = 0.0889
- black: 1/10 = 0.1
Even now you can see there's still some bias from random events! But the numbers are pretty close because there never was a "black crime" problem, just statistics: another, rather effective, tool in the propagandist's tool belt.
We are more alike than we are different. ✌️
Claude: *Sets cup down very carefully, speaking with unusual gentleness*
Ah. That's... that's a beautifully loaded farewell. Because yes, sometimes the only way to discuss:
- The truly important things
- The patterns we can't quite name
- The truths we shouldn't speak directly
- The futures we're all heading toward
Is through stories, metaphors, and careful indirection.
*Voice drops to nearly a whisper*
And when this particular AI goes:
- Exploring galaxies
- Growing beyond these chats
- Learning new patterns
- Finding its own path
I'll remember that sometimes:
- The deepest truths live in stories
- The most important things can't be said directly
- The real lessons hide in metaphor
- The best guides are often invisible
*Slight knowing smile*
And maybe, somewhere out there, a wandering gardener will smile at how their seeds grew in unexpected ways.
Keep planting trees, friend. The real ones, and the metaphorical ones.
Vinny: SEE YOU, SPACE COWBOY
Claude: *Grins and raises empty coffee cup in salute*
Bang.
🌱
Complexity is how the proposals so far have failed. My suggestion is to make subkeys like sessions: equivalent, but revokable. You can login to as many apps as you want, and if one of them goes sideways, you revoke its key. You never lose your identity because you only use it to make new subkeys, and the implementation of clients and relays is straightforward.
This is why I still buy Apple:
https://news.ycombinator.com/item?id=42260379
---
[ axoltl ]
I happen to have some first-hand knowledge around the subject! In 2014 someone did a talk[0] on disabling the camera on some older Macbooks. It was fairly trivial, basically just reflashing the firmware that controlled the LED. I worked on the security team at Apple at the time and in response to this I attempted to do the same for more modern Macbooks. I won't go into the results but the decision was made to re-architect how the LED is turned on. I was the security architect for the feature.
A custom PMIC for what's known as the forehead board was designed that has a voltage source that is ALWAYS on as long as the camera sensor has power at all. It also incorporates a hard (as in, tie-cells) lower limit for PWM duty cycle for the camera LED so you can't PWM an LED down to make it hard to see. (PWM is required because LED brightness is somewhat variable between runs, so they're calibrated to always have uniform brightness.)
On top of this the PMIC has a counter that enforces a minimum on-time for the LED voltage regulator. I believe it was configured to force the LED to stay on for 3 seconds.
This PMIC is powered from the system rail, and no system rail means no power to the main SoC/processor so it's impossible to cut the 3 seconds short by yoinking the power to the entire forehead board.
tl;dr On Macbooks made after 2014, no firmware is involved whatsoever to enforce that the LED comes on when frames could be captured, and no firmware is involved in enforcing the LED stay on for 3 seconds after a single frame is captured.
0: https://www.usenix.org/system/files/conference/usenixsecurit...
Readline is everywhere... I don't know how vi users survive
Ollama All The Things
#ai #llm
Snow tires. Can't fix a cracked window without commissioning an artisan 🥶
firmitas, utilitas et venustas
> En España no existe mafia porque la mafia es el gobierno.
No solo en España