The problem I run into is the propensity for it to cheat so you can't trust the code it produces.
For example, I have this project where the idea is to use code verification to ensure the code is correct, the stated goal of the project is to produce verified software and the daffy robot still can't seem to understand that the verification part is the critical piece so... it cheats on them so they pass. I had the newest Claude Code (4.6?) look over the tests on the day it was released and the issues it found were really, really bad.
Now, the newest plan is to produce a tool which generates the tests from a DSL so they can't be made to pass and/or match buggy code instead of the clearly defined specification. Oh, I guess I didn't mention there's an actual spec for what we're trying to do which is very clear, in fact it should be relatively trivial to ensure the tests match for some super-human coding machine.
All I hear about is how this is the 'shape of things to come' with regards to the AI bubble while nobody seems to care that France just told all the gov't agencies to stop using their stuff.
Losing out on EU governmental contracts seems to me to be somewhat of a big deal and the France thing is just maybe the first move in that direction.
They don't magically gain more privacy protection in public over what your average person has just because they clock out after a hard day of work by virtue of being a government employee.
They are constantly and consistently reminded that people have the right to record in public and they chose to ignore that as there are no consequences if they violate the law. Or that people have a right to peacefully assemble. Or freedom of the press...
I agree they don't gain more privacy protection in public than the average person. I also agree they shouldn't gain more privacy protection in public than the average public employee, either!
I'm merely assuming that the license plates being listed are ones they use for their official work, since the rest of their info is being tied to what's available for any other public work.
>> If someone ends up sniping a famous person, we go back in time and figure out who they are
Yeah, most civilians don't understand operational security at the functional level.
Though... most people doing these thing probably want to be caught because they just aren't quite right in the head and want people to tell them that what they did is 'justified' for whatever reason.
I'm going to go out on a limb and say because rust didn't exist 30 years ago?
Anyhoo... seems interesting. I've been trying to convince Claude to produce a verified JavaCard VM implementation, just for the hell of it, and this probably has a bunch of information to help with that.
Here's the problem, from what I've been hearing most of the actual criminals they've been "catching" are turned over by local and state law enforcement agencies with the rest are either in the process of criminal proceedings (that pesky "innocent until proven guilty" thing) or are involved in the immigration process as dictated by law.
They are literally pulling people out of judicial hearings, where the people are trying to comply with the law, and throwing them on airplanes without due process. Or just randomly snatching people off the streets with no probable cause including the occasional US citizen based on their (ancestors) national origin.
Seriously, my step-father's family became US citizens as a result of the Mexican-American War and the federal courts say it's probable cause to detain them based on their physical appearance. Like, WTF???
--edit--
Just remembered my grandmother saying she didn't teach her children Spanish because she didn't want them to grow up with and accent because she was literally beat if she spoke Spanish in school. True, this was 100 years ago but still...
No, blue state sanctuary cities do not turn over illegal alien criminals to ICE. They release them back into the community even though they have a ICE notice on them. ICE is then forced to track down these criminals themselves while being tracked and harassed by crazy far left agitators that do everything in their power to protect these criminals.
You know, for all my flaws I've always tried my hardest to be on the right side of history.
One can both believe that immigration policy is broken and also that the current way it's being enforced is immoral and unlawful. I took the same oath to the constitution when I joined the military as they did when they went into federal service and can see when when things are going off the rails.
My eulogy should be: "U.S. Paratrooper, Decorated Combat Veteran, Crazy Far Left Agitator"
IDK, it's probably more a matter where they don't want people to be flying RPGs into their windscreens and this is the first step for them to carry around frequency jammers. The last time I was in Iraq they used them to stop the cellphone detonated IEDs and all the convoys has one or two.
Coincidentally, folks won't be able to live stream their encounters but I'm sure that's totally unrelated...
>> Also note, i.e. stuff like statutory rape has been upheld even in cases where the perpetrator in all good faith thought the victim was 18+, the victim initiated selling the services, and the victim provided fake ID showing they were 18+.
You had me up to the "selling the services" part.
If you are 'engaging' with someone in a criminal enterprise it's probably reasonable to assume they might misrepresent certain facts to make the sale.
Yeah... I had to quit reading after seeing the No True Scotsman argument about who can and can't form an opinion on the state of the robots. I don't have any professional experience but feel like I can see enough, through experimentation using the daffy robots, to see what's happening but my only formal software education was when I was in high school in the 80s so I don't count.
And why do all these conversations center around web apps and whatnot? I don't work on anything like that but probably know a lot more about virtual machines and compiler theory than your average web programmer.
Been working on the design of an optimizer for my APL VM and asked Claude how this would fit in (as we don't have Dyalog extensions):
The pattern is:
f⌿ X ∘.g Y
Where f is an associative reduction and g is a comparison. iBurg could recognize this tree shape and apply rewrites based on operator properties:
┌─────────┬───────────────────────────────────┐
│ Pattern │ Optimization │
├─────────┼───────────────────────────────────┤
│ +⌿X∘.≤Y │ Binary search count (if X sorted) │
├─────────┼───────────────────────────────────┤
│ +⌿X∘.=Y │ Histogram / count matches │
├─────────┼───────────────────────────────────┤
│ ∨⌿X∘.=Y │ Membership (hash lookup) │
├─────────┼───────────────────────────────────┤
│ ∧⌿X∘.<Y │ All-less-than (single comparison) │
└─────────┴───────────────────────────────────┘
The generic rules would be:
1. Shape rule: f⌿X∘.gY avoids materializing n×m matrix if f and g have algebraic properties that allow streaming/early-exit
2. Sortedness rule: If X is sorted and g is monotonic (≤, <, ≥, >), use binary search
3. Associativity rule: If f is associative (+, ∨, ∧, ⌈, ⌊), can process in chunks or parallel
The cost model decides when O(n log m) binary search beats O(n×m) outer product -typically when both n and m exceed some threshold.
iBurg is the Bottom-Up Rewrite System (BURS) based optimizer to operate over the continuation graphs the parser spits out, not sure where the 'i' part came from though...
For example, I have this project where the idea is to use code verification to ensure the code is correct, the stated goal of the project is to produce verified software and the daffy robot still can't seem to understand that the verification part is the critical piece so... it cheats on them so they pass. I had the newest Claude Code (4.6?) look over the tests on the day it was released and the issues it found were really, really bad.
Now, the newest plan is to produce a tool which generates the tests from a DSL so they can't be made to pass and/or match buggy code instead of the clearly defined specification. Oh, I guess I didn't mention there's an actual spec for what we're trying to do which is very clear, in fact it should be relatively trivial to ensure the tests match for some super-human coding machine.
reply