I have Nexus 5, Xiaomi A1, Redmi Note 7, Samsung S7, and a Kindle Fire HDX, all running either LineageOS (Kindle) or PostmarketOS (the rest). PmOS ones run some not very demanding containers (scrapers) on k3s.
Or simply we use AI and see on the ground what it can and can't do. I can generally trust an agent for solved problems, but the more something deviates from established industry standards (i.e. what was relentlessly scraped) I have an increasingly harder time not having constant oversight of what it's doing, no matter the specs I put on the md.
Personally I feel most of the improvement in the last year comes from tooling/integration (MCPs, realtime documentation access, treesitter support, orchestration) than from the models themselves, in the last year. And still frontier models would routinely come up with bs until you tell them to actually use those tools.
You're talking as if this is a static thing though. It's the God of Gaps [1] but for humanity's special sauce.
Two years ago, I couldn't trust an LLM to do anything that wasn't straight forward boiler plate.
One year ago, I was pretty solid at writing algorithms that were combinations of existing ideas.
Now, Fable is outputting stuff that I would genuinely consider to be creative and original if a colleague had presented it to me.
Yes, maybe the code style still isn't great, but given the pattern of the last few years, it feels correct (a priori) to assume that this gap isn't going to keep closing.
I wouldn't call it two-tier justice, because generally the courts do the right thing, but there's a shamefully obvious two-tier policing.
From the Jay Report [0] showing crimes swept under the rug according to ethnic/socioeconomic background of perp and victim, to arresting people for opposing genocide (sorry: terrorism!) [1] to the recent case of Henry Nowak [2], it's really hard not to see a two-tier policing in the UK. And this very submission; caring about privacy is seen grounds for being reported and potentially investigated, by a private company! Which suggests it's something already internalized, too, for people who resist big corp surveillance.
Back in the 90s and before, the two-tier heavily punished the minorities, and in an overshooting overcorrection, now it's the other way around. Nowak getting handcuffed by cops going "I don't think you have [been stabbed] mate!" says it all. Unless it's regarding opposing/supporting Israel, then the two-tier flips and people with basic human decency and actual antisemites are pigeonholed together, nevermind their background.
The Palestine Action ban was overturned by the courts. And putting this together with the Nowak case leads to an almost contradictory position. Whatever the issues with police handling of pro-Palestinian protestors, they're not instances of what the far right would call two-tier policing (i.e. the bogus claim that the police treat white people more harshly than other ethnic groups). For example, here is a poster who' all over this thread complaining about two-tier policing, but who supports the ban on PA: https://news.ycombinator.com/item?id=48155504
The more nuanced reality is that the police are extremely imperfect and treat all sorts of different groups of people unfairly in all sorts of different ways. That's indeed a problem, but it's not a political conspiracy.
That's because there's a pending appeal. It may take a while for the situation to fully resolve, legally speaking, but I think it's unlikely that the ban will be sustainable in the long term.
How would you propose to immediately put an end to the bad thing? Checks and balances take time. Any system that allowed this to be immediately fixed would also allow a whole bunch of other things to be immediately fucked up.
>The Palestine Action ban was overturned by the courts.
Which is exactly what I said, courts generally do the right thing. The people were still arrested, and the dissent was still crushed, though. Israel is a strategic partner so all protests are equal, but some are more equal than others.
>they're not instances of what the far right would call two-tier policing (i.e. the bogus claim that the police treat white people more harshly than other ethnic groups)
While far-right popularized the term, the awareness is now, at least among the people I know, full-spectrum. The tiers are independent according to the issue at hand and the priorities of the powers and lobbies involved. Simply some camps were unaware until it was their turn, and become confused when they overlap.
>The more nuanced reality is that the police are extremely imperfect and treat all sorts of different groups of people unfairly in all sorts of different ways. That's indeed a problem, but it's not a political conspiracy.
Personally, I think that statement has a pass in many places, but not in the UK. The "policy projection of values" (to give it some name), with the growth of identity politics mixed in to make it worse, attempted to make everyone feel equally protected, but instead succeeded in convincing almost every demographic in the UK that the system is rigged against them.
We are not racist -> certain crimes by given little publicity or ignored depending on the background of who committed them. Two tier.
Israel is a strategic partner -> pro-palestinian/anti-zionist groups labeled as terrorists. Two tier.
Hate has no place here -> police dispatched for nasty online comments, but dragging their feet for actual physical crime. The fact that "Non-Crime Hate Incidents" is an actual term is just baffling. Two tier.
Net Zero -> right wingers complain about blockades of ecologists going unpunished. Two tier. Police, Crime, Sentencing and Courts Act 2022 and the Public Order Act 2023 passed, ecologist arrested (and sentenced!) for protesting against not doing enough for Net Zero. Pendulum swing, still two tier.
Directing mind and will -> individuals are fully accountable by their actions, for corporations one need(ed) to prove "directing mind and will" of the executives. Two tier. See the British Post Office Horizon scandal for extra depression, or how individuals are prosecuted for dumping, yet Thames Water can dump inordinate amounts of sewage but that's just "regulatory failure". While recent acts (Crime and Policing, Failure to Prevent Fraud) are in the right direction, criminal penalties for corporations remain financial, and swallowed into operating costs. Still solid two tier.
Russia is a criminal state -> Londongrad unbothered, alive, moisturized, flourishing. Two tier.
Protect the NHS -> partygate for some, fines for others. Two tier. Don't get me into the defunding of it.
Peace above everything -> the state actors that committed crimes (often acts of terrorism) with impunity in Ireland during The Troubles, were later prosecuted, while IRA members were secretly given out-of-jail cards. Just compare John Downey vs Dennis Hutchings. Since the topic has cooled down and the Hutchings' case fueled the fire, they passed the Legacy and Reconciliation Act to, 25 years later (!) finally providing immunity to British criminals. This is particularly damning because those following orders had been prosecuted, but no single commander, general, intelligence officer, minister actually giving or overseeing those orders had been, at any point. Two tier.
The UK is in a permanent state of "fake it till you make it" of the idea of the perfect state, and in doing so it routinely over/undershoots the mark, in an endless Samsara of overcorrections, because it is all still fake. The state sets the target, but it does nothing to get there. Ultimately the consistent application of the law is second to state, corporate and geopolitical interests, meaning the two tiers will be there no matter how correct the laws are.
Your comment perfectly illustrates that all sorts of groups of all sorts of different political stripes (or none at all) have been treated unfairly at times. All of the issues you raise are serious ones, but they don’t support the spurious narrative about “two tier policing” being pushed by Musk, Farage and their ilk.
But you can tell it to once (in CLAUDE.md for example) and it will nearly every time (it's getting much better at that). Since opus 4.7 (which I consider a downgrade overall) it's been much better at following CLAUDE.md . I even have an intentional contradiction in my user-level CLAUDE.md and the project levels, so I can tell which one is taking precedent or if both are disregarded, and it follows at least one of them most of the time, and it follows the local one 95% of the time.
While they absolutely do fail as you say (though in my experience not by default), this failure mode is still a massive improvement over the frequent human case of guessing based on the function/class/property/argument names.
Now, a really good human collaborator who reads all the stuff and thinks carefully, that was still better than what I saw from AI models at the start of this year. But I've also worked with my share of idiots, and been one too.
I'm not going to get into if *current* models can or can't reliably do any particular thing to any particular standard; previously my comparison was the same conversations with regard to video game computer graphics in the 90s always being "photorealistic" when they really weren't*; now, I'm starting to feel such discussions have the same vibes as Tesla fans insisting that "FSD-{insert current version here} solves all the problems and is a real breakthrough and the Rototaxi will totes conquer the marketplace this time for real bro, just one more version bro", etc.
if you find yourself saying 'if you tell it to' a lot about LLMs that usually just says something about your prompting methods.
or, in other words , if you want the thing to always read the documentation then make that a strongly highlighted point both in pre-prompts, active prompts, and memory.
It mustn't always nor never. It should follow a best judgement based on the .md, toml or whatever you use; in the end it's up to the LLM to decide which registered tools/mcps are used, and if the LLM is confident about some bs it will use that confidence instead of the tool.
When people complain about it, it's more often a gap between different knowledge domains and hard to measure characteristics of the environment, than it is an actual "you're using it wrong".
I'm on the other end, Gmail sends to spam all sorts of legit things. Including mails from the "Google Assistant Privacy Litigation Settlement", conveniently enough.
.001 for creatine levels isn't surprising; that's a lot of creatine. I'd explain the cognitive tests with the practice effect, because it is unlikely that creatine had such a massive effect and we only discover it now.
I hear about tech bros taking creatine these days with the tone of voice that they use to talk about microdosing. So I can’t imagine it having zero effect.
What I worry about more is that it has more to do with fixing a deficiency. That being deficient in creatine causes a cognitive loss more than supplementing causes a boost.
As someone who's microdosed (though mostly normal-dosed and occasionally megadosed) in the distant past, the whole microdosing fad was equal parts entertaining and baffling. Anecdotal, but from all people I know that has taken psychedelics, only one doesn't find it to be a waste.
Maybe a better analogy would have been the Balmer curve. I wasn’t trying to imply psychedelics are unhelpful, just that we should be careful of suggesting coding productivity gains while on them.
Also IMO, the Balmer peak is a stronger effect than creatine.
I imagine it's due to having had decent enough GPUs and decent enough CPUs, from a single vendor.
If you want the platform to be x86 but not AMD then your only other choice is Intel, but they've only recently started making high performance GPUs. So then you need another vendor for the GPU, and your only choice is Nvidia.
A lot simpler, cheaper and predictable to go with a single vendor for both I imagine?
AMD also had the strongest offering for GPU and CPU using the same memory with the same address space. That allows you to switch between CPU and GPU processing for the same data, without paying the cost of moving the data to and from the GPU. Similar to what we now have on Apple silicon
They tried to push the same into the desktop market with their APUs, where it was mostly ignored. But console games only target a couple hardware configurations, making it viable to take advantage of such hardware features
Also also, AMD’s play has always been to produce HW that offers good performance/$, with the downside of having much weaker SW offerings to go with it.
Consoles are always pressured to minimize upfront purchase costs, and they generally replace the vendor-provider SW stack with their own anyways.
And they’ve been in a rough spot at various times in the past, which probably made them willing to negotiate with the console companies.
Actually looking at this thread, there’s a lot of good reasons they were the go-tos for consoles. Consoles seem to be in rough shape at the moment, I wonder if part of that is that AMD has been doing too well since Zen, haha.
You’re approaching this as if every company had the same corporate intentions.
Nvidia never cared much for those types of deals. They preferred to lose Apple as a business than to admit fault, they’ve always refused to compete on price for the business of Sony and Microsoft’s consoles. They’re adamant to beat at the sound of their own drum.
reply