m11a

8 ms·

61.
by m11a 1y ago
The solution is kagi.com imo. Before AI generated results, the first page of Google was SEO-optimised crap blogs. The internet has been hard to search for a while.
62.
by m11a 1y ago
Indeed. And aside from that, LLMs cannot generalise OOD. There's relatively little training data of complex higher order constructs in straight assembly, compared to say Python code. Plus, the assembly will be target architecture speci
63.
by m11a 1y ago
I feel like they’re trying to increase switching costs. eg was huge reluctance to adopt MCP and each had their own tool framework, until it seemed too big to ignore and everyone was just building MCP tools not OpenAI SDK tools.
64.
by m11a 2y ago
> Do we suppose they will just create a backdoor to allow _some_ bots in? That, and maybe they will as CF seem quite big on MCP.[0] Or people just bypass the bot detection. It's already not terribly difficult to do; people in the sn
65.
by m11a 2y ago
Neither, my categories only cover "services" (at least as Wikipedia would categorise things into this bracket: https://en.wikipedia.org/wiki/Service_economy ). I agree with you on construction and physical wor
66.
by m11a 2y ago
You'll be happy with either. Bishop's approach is historically more mathematical (cf his 2006 PRML text), and you see that in the preliminaries chapters of Deep Learning, but there's less of this as the book goes on. I'v
67.
by m11a 2y ago
Most ML is disjoint from the current NN trends, IMO. Compare Bishop's PRML to his Deep Learning textbook. First couple chapters are copy+paste preliminaries (probability, statistics, Gaussians, other maths background), and then they co
68.
by m11a 2y ago
If so, then it seems the solution is for HN to upvote the random qualified engineer with AI predictions?
69.
by m11a 2y ago
"stuck" is a bit strong of a term. 6 months ago I remember preferring to write even Python code myself because Copilot would get most things wrong. My most successful usage of Copilot was getting it to write CRUD and tests. These
70.
by m11a 2y ago
> that's not what _humanity_ is about I've not spent too long thinking on the following, so I'm prepared for someone to say I'm totally wrong, but: I feel like the services economy can be broadly broken down into: ple
71.
by m11a 2y ago
Syntactically it's a pretty nice language, with a nice and sane ecosystem (crates, etc), and it's fun to write (which is a pro especially for unpaid hobby projects). If you get used to it, it's nice to write various things in
72.
by m11a 2y ago
I mean, if it was solely about renting machines, we’d all just use DigitalOcean, or EC2 on AWS. People use things like RDS and EKS/GKE to avoid all the administrative overhead that comes with running these things in prod. The database
73.
by m11a 2y ago
I’d like to see some thoughts on where we go from here. Is there a way we can keep end users protected even despite potential compromise of services like ToDesktop? (eg: companies still hosting some kind of integrity checking service themse
74.
by m11a 2y ago
It's still my favourite web framework. I just wish the Ruby language had better support for type annotations (like Python does). Then it'd be sorta perfect for me
75.
by m11a 2y ago
I agree. But what I mean is: that's not how it's perceived in the current interview structure, which lasts maybe 45 minutes or so. Ultimately, going down a dead end means you'd now have 30 minutes to find the right solution a
76.
by m11a 2y ago
I dislike pair programming interviews - as they currently exist - because they usually feel like a time-crunched exam. You don't realistically have the freedom to actually think as you would in actual pair programming. i.e. if you wa
77.
by m11a 2y ago
> I am curious to know more about the limitations / perils of using LLMs to train/evaluate other LLMs. At the extreme, there is this paper on ‘inbred LLMs’: https://www.nature.com/articles/s41586-024-07566-
78.
by m11a 2y ago
> In fact there isn't a single widely used language besides Java that feature checked exceptions. Not in Java's form, but I'd argue Rust's Result type basically achieves the same effect but with less syntax. A functio
79.
by m11a 2y ago
One example from today: I had a coding bug which I asked R1 about. The final answer wasn't correct, but adapting an idea from the CoT trace helped me fix the bug. o1's answer was also incorrect. Interestingly though, R1 struggled
80.
by m11a 2y ago
A lot of the niceness about DeepSeek-R1's usage in coding is that you can see the thought process, which (IME) has been more useful than the final answer. It may well be that o1's chain of thought reasoning trace is also quite goo
81.
by m11a 2y ago
Huh? China has political censorship by law: https://en.wikipedia.org/wiki/Censorship_in_China It’s much harder if anything, due to the sheer amount of topics censored.
82.
by m11a 2y ago
Well, if we’re using the output of AIs, writing blog post articles with it and using AI to post comments on Reddit/X etc (as some do), and a few years later OpenAI et al refresh their datasets to train a new model, then you’re doing ex
83.
by m11a 2y ago
In a typical British parliamentary system, the executive also has majority in Parliament. If the executive doesn’t have parliament, they lose the executive. ‘Impeachment’ in Parliament systems only works when MPs are willing to think for th
84.
by m11a 2y ago
I much prefer Rails to say Django, but I wish decent typechecking were possible in Ruby. All the options I’m aware of (Sorbet, header files, etc) are all so inferior to Python’s typing syntax that I feel like this alone makes me not want to
85.
by m11a 2y ago
But Rust is a systems language, for which Rust is excellent. Among its peers, I find C++ blog posts even less coherent. Despite recent trends to use Rust for everything, it’s not supposed to be used for CRUD apps. I think this blog post is
86.
by m11a 2y ago
> Base models are uniquely valuable as a research product because they accurately model the distribution of the data that they were trained on, as opposed to models that have had substantial RL tuning done to collapse their generation di
87.
by m11a 2y ago
But then you’d be holding a DB connection for the entire duration of your task (which may include HTTP calls, etc). You might even do asynchronous work in parallel, which doesn’t quite work with txn locks. So the session based locks seem a
88.
by m11a 2y ago
You don’t need to be a great engineer for most startups, unless they’re particularly specialised. Most apps are really just CRUD at small scale. These apps aren’t really hard until you scale. Eg Airbnb is really simple to make, technically
89.
by m11a 2y ago
Risk of bad rep when the researcher reports to HN or makes some noise. Then future security researchers don’t try find issues on your platform, and it’s more insecure as a result. For a sensible large company, it’s not worth being stingy ov
90.
by m11a 2y ago
> Based on what this person was looking for (someone to build out his ideas) he would be better off trying to get funding and hiring someone. Yeah.. or even just learning how to code himself. Could learn a decent amount in 9 months.

More ›