Presumably the police computer system is told about the transfers automatically and it has no capacity problems. I know my own country has long had such a system and these days is almost entirely cashless.
2,072 karma · joined February 18, 2023
And, as the professor in the article explains, it is a “disaster” for a small minority. And those are not he same minority who struggled for the same reasons before, and those difficulties ought be addressed. The system can be improved.
But it’s largely a success for the vast majority too. I don’t personally know of anyone with a negative impression of it. It’s actually something that the average Nordic Baltic person is so used to and happy with that it only comes to mind when we meet people from countries that aren’t organised - and we feel sorry for them!
It’s the same situation with cash. Very few people in a cashless society are wanting to go back to the old ways.
First, the reason LLMs learned to like em dashes is that they are common in the training corpus - they are a thing before LLMs that LLMs have learned, not invented?
Second, work browser has nice blue swiggles under everything I write into a textbox. I dutifully click through them and accept the rephrasing suggestions. I get a lot of em dashes. My blog posts and whitepapers and stuff are full of them and other “AI tells” - but I think they read better because of it.
That they don’t suggests that really it is only incrementally better than Opus 4.7 and that the market won’t bear a price increase that makes it economical to serve let alone profit from serving.
So the cynical me imagines execs sitting around the table and worrying that releasing it at anywhere close to break even would risk actually hurting the brand instead of setting them up as a premium company, and this at a time just before ipo when they can ill afford that rumour.
So they wonder what to do, and think playing national security card is the obvious way out. It’s incrementally better enough to find bugs that previous sota missed, it doesn’t get used widely so it’s cheap to serve and they get the good publicity without the economic scrutiny?
Making a loss selling to a small number of users using it in a limited way is entirely affordable. Making a loss selling it at scale is correspondingly unaffordable?
Are there any much more detailed walkthroughs of how it works and how it decides the tools to use and the grep to use etc and what the conversations actually look like?
In the UI you see just enough to know it’s doing something but you don’t really see the jumps it’s making offscreen.
The aspect I miss is the distributed compilation hinted at in the article. I remember back at the end of 1990s using distcc and things, but that never seemed to happen in the Java world and the tooling like maven etc is structured to make everything one long dependent chain. Shame.
But yeah for me Java took a big leap forward when it got lambda but there hasn’t been another big leap since. Still so many obvious features and sugar to add to file off the sharp edges and remnants of enterpriseness.
Ddos protection services can be cast as a digital protection racket where they have a perverse incentive to keep attackers attacking. “It's a dangerous internet out there; you'd better pay us to protect your website from the attackers using our free tier.” At the least, even if there is no active collusion or profit sharing or anything like that, there is not a clear side that the DDos protector service is on?
Does code extraction and manipulation fit in that? Would interfaze be the agent that a coding agent uses?
In the same way you can imagine the Chinese government pushing the release of deepseek etc to make sure no one thinks the US has “won” and to keep everyone aware that a foreign model might leapfrog in the short term future etc.
At some point though if OpenAI/Antropic/Google plateau or go bust then the open source sponsorship becomes less likely, as making it open source was a weapon not a principle.
Years ago, I entered a Scrabble programming contest and needed to compress a GADDAG dictionary to fit into my 6MB L3 cache. Without knowing the official name for it, I ended up using the exact same suffix-compression mechanism by moving characters to the edges instead of the nodes to merge overlapping paths.
Sharing my old write-up here in case you or other data-structure nerds find the overlap interesting! https://williame.github.io/post/87682811573.html
I put my stuff through several sota models and round robin them in adversarial collaboration and they are all useful even though, fundamentally, they don’t “understand” anything. But they are super useful delegates as long as deciding on the problem and approach and solution all sits safely in your head so you can challenge them and steer them.
So I know the article is about one particular new model acing something and each vendor wants these stories to position their model as now good enough to replace humans and all other models, but working somewhere where I am lucky enough to be able to use all the sota models all the time, I can say that all keep making obvious mistakes and using all adversarially is way better than trusting just one.
I look forward to the day one a small open model that we can run ourselves outperforms the sum of all today’s models. That’s when enough is enough and we can let things plateau.
I start with a high level design md doc which an AI helps write. Then I ask another AI - whether the same model without the context, or another model - to critique it and spot bugs, gaps and omissions. It always finds obvious in hindsight stuff. So I ask it to summarize its findings and I paste that into the first AI and ask its opinions. We form an agreed change and make it and carry on this adversarial round robin until no model can suggest anything that seems weighty.
I then ask the AI to make a plan. And I round robin that through a bunch of AIs adversarially as well. In the end, the plan looks solid.
Then the end to end test cases plan and so on.
By the end of the first day or week or month - depending on the scale of the system - we are ready to code.
And as code gets made I paste that into other AIs with the spec and plan and ask them to spot bugs, omissions and gaps too and so on. Continually using other AI to check on the main one implementing.
And of course you have to go read the code because I have found it that AI misses polishes.
Would there be any value in using simd to check the whole cache line that you fetch for exact matches on the narrowing phase for an early out?
There are three tiers of discount. There are general special offers, advertised widely, to get people through the door. Then there are members only discounts, advertised in store, to get people to be members. Then there are mailed to members personal discounts that are uncannily accurate for what is running out to get you to go shopping now instead of later. These days the hand scanners and apps also tell you what coupons you have so you don’t need mail but junk mail is still regular.
TBH as a shopper it works great.