crazylogger

5 ms·

1.
by crazylogger 15d ago
Let's say Claude Code's system prompt is updated to recommend responding in Simplified Technical English. It'll work for the current generation of models. But Anthropic will train the next generation via RL on Claude Code tra
2.
by crazylogger 19d ago
Assuming there is an index, agent still needs to write queries against the index to get any information from it - might as well grep!
3.
Don't Let LLMs Play Telephone with Your Ideas(blog.yfzhou.fyi)
3 points by crazylogger 21d ago|0 comments
4.
by crazylogger 1mo ago
Anthropic server decrypts it as part of fulfilling every request, and haiku recites it per your request.
5.
by crazylogger 2mo ago
It's not about discrediting AI. We know LLM is a commodity technology like electricity at this point. If somebody in 1900 claimed they had a setup at home where they feed in electricity and cool air comes out the other end (meaning the
6.
by crazylogger 2mo ago
They should just be symlinks to CONTRIBUTING.md and other docs humans are meant to read. The whole idea behind LLM is they use the same language and tools as us, so we literally don't design anything specifically for them.
7.
by crazylogger 2mo ago
All the Chinese reporting I see point to the second (majority) interpretation. Liang being furious about his private investor talk leaked online is the news here. e.g. https://x.com/_FORAB/status/208103450010101761
8.
by crazylogger 2mo ago
The goal is to test for AGI where G stands for general, that means ability to act in any environments, ideally solving novel tasks using novel tools we’ve never seen before in the world. If a specific prompt or tool design lifts a model’s
9.
by crazylogger 2mo ago
Certainly sounds like a "P vs NP" style conjecture that shouldn't be possible in practice, save for certain generalizations, such as "this is a cybersecurity task, we know Fable will refuse (and score zero), so we just r
10.
by crazylogger 2mo ago
> Oracle routing is a method for measuring the best theoretical performance by running the task through each model and then picking the cheapest correct option (the cost/performance ceiling). Their "router" is an oracle re
11.
by crazylogger 2mo ago
The Y axis is total cost in USD. For it to go down would mean OpenAI refunding you money.
12.
by crazylogger 2mo ago
Manufacturing is primarily in China - that's true for Go Pro & everyone else and almost needless to say. The point is China usually eats the design layer too, making Apple a little unique in that they survived Chinese competition c
13.
by crazylogger 2mo ago
Even ignoring chip export ban, Chinese companies have way less funding than American counterparts, maybe 1 or 2 orders of magnitude less depending on which company you look at. Deepseek’s recent big funding round being “only” a couple billi
14.
by crazylogger 2mo ago
To be fair while almost no company publishes what a "credit" or what a 5h window in their subscription plan really is in terms of input/cached input/output tokens, Xiaomi does: https://mimo.mi.com/docs&#x
15.
by crazylogger 3mo ago
Students will just one-shot whatever lab problems professor comes up with using Claude Code.
16.
by crazylogger 3mo ago
Actually they don't even need to compete against frontier closed models, they just need to work. 99.99% people's day jobs aren't competing for the Fields Medal or even finding security vulnerabilities. So it appears while TAM
17.
by crazylogger 3mo ago
In a typical agent loop your N-th LLM request naturally becomes prefix for the (N+1)-th request. As the thread grows longer, cache hit rate converges to 100% and unit pricing for cached tokens is 10-100x cheaper.
18.
by crazylogger 3mo ago
Chinese labs access Claude via API. Isn't it the black box method by definition?
19.
by crazylogger 3mo ago
Xiaomi Mimo v2.5 is my favorite alternative. Matches DS v4 Flash (official) pricing exactly and supports image/audio/video input.
20.
by crazylogger 3mo ago
Cache hit rate dominates your total cost calculation for long agent session, and it largely depends on the provider. Deepseek's native deployment is probably much better than third party in this regard. For v4 pro it's a whopping
21.
by crazylogger 3mo ago
The promise of intelligence might be larger still. By scaling and using superintelligent LLMs to write code for itself, it's possible that the whole field of robotics is just another problem you can point LLM agents at and expect to be
22.
by crazylogger 3mo ago
This is equivalent to buying 3 extra days of free time with 60% of your income. You want more (holidays, in this case) and you buy them with work.
23.
by crazylogger 3mo ago
They basically said "Deepseek ran 150,000 requests and here's the gist of one of their prompts". Anthropic doesn't know which accounts are Deepseek proxies beforehand, so definitely sounds like retrospective analysis of
24.
by crazylogger 3mo ago
Didn’t they all but admit they’ve been storing and actively looking at requests with this post: https://www.anthropic.com/news/detecting-and-preventing-dist... ? If they weren’t storing, they’d be oblivious to what cus
25.
by crazylogger 4mo ago
> It might seem weird for us to teach beginners Python, knowing that they’ll then have agents output other, faster languages. I see an analogy here with Chinese: Many people don’t realize this, but children in China first learn Latin cha
26.
What we learned building sandbox for document agents(blog.yfzhou.fyi)
1 points by crazylogger 4mo ago|0 comments
27.
by crazylogger 4mo ago
OpenRouter may see you fire hundreds of requests at them, but they have no idea that "these 50 requests here at 4PM are for task A", "those 100 requests there does task B", etc. So it's a shallow analysis at the &qu
28.
by crazylogger 4mo ago
For a single database using UUIDs, yes, it's astronomically rare. But it's quite a different thing to say that no computer system on Earth has ever experienced a UUID collision. The number of systems out there is also astronomical
29.
by crazylogger 5mo ago
People had this "why you probably can't run a GPT-4 (or even GPT-3.5) class model on your MBP anytime soon" conversation before. Today's LLMs are able pack much more capabilities into fewer parameters compared to 2023. W
30.
by crazylogger 5mo ago
"Hi ChatGPT, propose and prove something radically new in the genre of Gödel's theorem." How is this not just another proposed problem (albeit with a search space much larger than an Erdos problem's)?

More ›