The cross-correlation function can be applied to measure time shift between two waveforms, even those collected from different devices. The accuracy of this is not limited by the sample period. In fact, the two waveforms need not even have a common sample period. I think this approach could let simonw's network become self locating.
The problem was actually described in the agent output that I hadn't read when requested that additional feature:
> Distances only fix the layout up to translation, rotation and mirror image. That doesn't matter for the maths, but the map I'm shown has to match the real room, so the UI needs "rotate" and "flip" controls, or a way to say "Blue is next to the door".
My biggest problem is remembering what I did yesterday. It's starting to become hard to remember what I have going on right now. Multitasking is a killer.
> Search is so bad and LLM so good that it's faster and easier to re-implement than reuse.
This is a scary level of AI mania.
Even if you're being assisted by the best frontier model, there is no way that reimplementing a PDF engine from scratch is faster than just using the one that's already there. AI-driven coding still takes time and tokens. A web search takes maybe 60 seconds to search and find the existing implementations?
You could even tell the LLM to configure (or even fork) one of the existing implementations for you, and that would still be way faster than a from-scratch implementation.
LLMs will miss a lot of edge cases. This is critical for a library that parses PDFs. Ask it to review the code for issues and by the 20th time you ask, it still will find issues. How's that faster (or safer)?
I use AI heavily but it's not this omnipresent deity that knows everything at all times.
Am I alone in thinking type hints are largely a waste of time in Python and can be actively harmful?
I had claude run ty via this new "uv check" on a moderately sized Python package (cli+tui to a multi-transport messaging system) and it found 67 things. Claude checked each finding for bugginess and found 1 is marginally worth fixing. A large fraction of the findings were ty guessing the type and being wrong. Hinting would I guess remove those wrong guesses but hinting is a PITA and greatly increases code size, both of which is bad for human and LLM.
Do people really get benefit from going all-in with type hinting?
Type annotations are generally great, if you've never had the displeasure of working with Python pre type-annotations, you'd know that tracking down "wtf type even is this" was a big part of just working in Python. The extra text/tokens spent on annotations are 100 % worth it, both for humans and LLM.
An issue you run into these days is that even if you have excellent type checking on your own code, some libraries don't. You of course want to avoid such libraries, or see if you can find type stubs. Gradual typing can suck, but well-typed Python is a bliss.
I work with Python today without type annotations, don't find that kind of type tracking difficult at all, nor do I find the annotations worth the effort. The questions I've seen over the years, meanwhile, from people trying to force the annotation system to do strange things, have suggested to me that people are creating far too much complexity for themselves. Incidentally, IMX, ChatGPT usually writes un-annotated new code unless you specifically ask for annotations. And it can easily refactor that un-annotated code, too.
Annotations are ignored by the compiler (except insofar as they require adding some attributes to objects) and do not guide the bytecode-compilation process or empower any kind of optimization (without a dedicated third-party system like mypyc). Nor are they relevant to PyPy. They are not even free abstractions: as mentioned some objects gain additional attributes (in case you want to introspect something at runtime, but you usually don't), and e.g. calls to `typing.cast` still have to be looked up and made even though it's explicitly intended as a no-op. Plus, expecting annotations to reflect types that can be statically checked gets in the way of other interesting uses of annotations (and the original PEP explicitly designed with other use cases in mind).
I have been using Python for over 20 years. If anything, my attitude has become further against type annotations in the last few.
I routinely run into issues, where Claude will try to add functionality to a dict-heavy API and messup the keys, I just do an initial pass converting everything to dataclasses, and it can iterate on it's own with the typechecker to fix everything on it's own, but otherwise I have to do a bunch of manual iterations.
The other great thing about type annotations is that you dont have to use them. Coming from strongly typed language background its life changing for me to sometimes rely on them sometimes just ignore.
Dude, the number of people who disable javascript is a rounding error compared to those who don't. Even in the corpus of 'Hacker News readers', I'd expect it to be miniscule.
It might be one of the things you personally care deeply about, but that doesn't mean any site owner needs to. At all.
Or, wrong by design. There is something about nudging users to use the claude CLI from which Anthropic extracts value. Otherwise they would not charge extra for "claude -p" or ACP access.
The tmux send-keys + terminal scrape is the LLM version of the analog hole of DRM'ed audio.
Wait, is `claude -p` usage counted differently? I've built something of a multi-agent daemon system and use `-p` extensively. Now I'm wondering if I'm burning my credits faster.
If you are paying for the API anyways, I don't think it makes a difference. What is different is if you are on a subscription then you need to pay more to use `claude -p`. But then Anthropic started adding free credits, at least for Team and above plans, to offset the cost. Drug dealer pricing.
But don't worry, I'm sure next week it'll all change in some new and confusing way. And not for the benefit of the customer.
Money for the Genesis Mission is coming out of science budgets. Given how damaging this funding for Artificial Intelligence as been for US science, can't wait to see what Super Intelligence funding does to it.
reply