nullbio

14 ms·

301.
by nullbio 2mo ago
It's caused by human laziness and corner cutting. LLMs can write buggy code or incomplete architectures as much as humans can, but standards have lowered. It's not the LLMs are not capable of also fixing these same issues, but tha
302.
by nullbio 2mo ago
Competitors can reproduce your discernment as well, it's the sum of your product and can be equally cloned. So how is this a moat?
303.
by nullbio 2mo ago
Perfect for consumers. You buy it and then you need to buy a new one in a couple of years. If they can make them affordable they'll sell like hotcakes.
304.
by nullbio 2mo ago
Yeah, but that's boring. :)
305.
by nullbio 2mo ago
These capabilities that emerge are not "orthogonal leaps". They are very much just more of the same thing.
306.
by nullbio 2mo ago
I think LLMs perform better at code than human language tasks because there's no clear way to eval human language tasks in a non-ambiguous or concrete manner. Any sort of eval that happens around language is transformation tasks, which
307.
by nullbio 2mo ago
How on Earth do we solve this bloat and death-by-a-thousand-cuts issue with frontier LLMs? Are there any actual solutions or attempts at solutions to this problem that I can try? Any tools or frameworks? I've tried re-architecture skil
308.
by nullbio 2mo ago
It's easier to skim pseudocode than it is to reason about complex logic. When I say pseudocode I'm talking very high level but structured explanations, and less the sort of traditional pseudocode you'd imagine which is basica
309.
by nullbio 2mo ago
It also requires no form of ID, age verification, or even an email address. Furthermore, the bad porn on Reddit is far more heinous than anything on any regular porn site. Stuff that shouldn't even be legal - and may not be. So that&#x
310.
by nullbio 2mo ago
Definitely not. The motive is because it's easily gamed for propaganda. It's the easiest platform to bot, and thus exploit, for psychological manipulation. Text is also the cheapest way to influence people - making videos is much
311.
by nullbio 2mo ago
That would be nice. Unfortunately people would just end up LLM generating slop filler descriptions anyway though.
312.
by nullbio 2mo ago
I (and I imagine many others) would love to use something like this, but can't, because my data is too sensitive to be uploaded to a cloud of which I have no gaurantees of privacy/security. Is there any way we do this using rented
313.
by nullbio 2mo ago
Yes, and it's only logical things move in this direction because there's massive hardware incentive to do so. If frontier models can be broken down into small models, networks of smaller-GPUs can be utilized. Right now the smaller
314.
by nullbio 2mo ago
> The idea is that we get specialized models that are better then general purpose models. But its rare for a specialized model to beat a strong general model. Are you sure about that? I mean, MoE is basically an array of specialized mode
315.
by nullbio 2mo ago
Yep. It's a real shame that the labs are incentivized not to go in this direction. They all want to try and suck us into the cloud and take away full control and local processing, but there's far more opportunity by building small
316.
by nullbio 2mo ago
I've seen many of these pop up over the last 6 months. What separates yours from all of the others? What makes it better? (Genuine question, not being snarky.)
317.
by nullbio 2mo ago
So basically a layer built over Jujutsu?
318.
by nullbio 2mo ago
Good luck. People can't figure out how to solve ARC-AGI reliably, let alone the complex problems in the real world.
319.
by nullbio 2mo ago
A better interface for code review would be a pseudocode layer that sits on top of all of your code, at the editor level. Basically it would convert all of the chunks of your code using an LLM to simple pseudocode with clickable symbols so
320.
by nullbio 2mo ago
The fact that they'll ban kids from accessing every social media except Reddit tells you everything you need to know about the real motives behind this "ban".
321.
by nullbio 2mo ago
Go really feels like it was build for agentic development. Surprised it's not more popular than Rust.
322.
by nullbio 3mo ago
Yeah, which is ironic considering their track record too. The most recent one is every shared Claude chat has been indexed by Google and is searchable on the search results. Then there was the RCE in their CLI that they had for a year and w
323.
by nullbio 3mo ago
"Open-weights models that don’t have dangerous capabilities are a public good" Read between the lines folks. Anthropic deems every model that has frontier capabilities as "dangerous", and thus they are against them. We
324.
by nullbio 3mo ago
This really means nothing. You can ask the same thing in Chinese to Opus and it will tell you it's DeepSeek, for example. Things leak into the training data and models hallucinate.
325.
by nullbio 3mo ago
"Then we'll be able to guesstimate if "labs are subsidising tokens on API pricing"." - The thing is though, that Anthropic and OpenAI have breakthroughs in optimization that none of the Chinese labs have. So it&#x27
326.
by nullbio 3mo ago
ast-grep is a great tool. Really helps with agentic work. I wish the LLMs were trained to use it by default. Would make them incredibly powerful.
327.
by nullbio 3mo ago
The behavior of the two companies and their culture are entirely different. OpenAI listens to their audience, they engage with the community, they're constantly resetting account limits for customers and admit when they mess up, they&#
328.
by nullbio 3mo ago
Not sure why people keep lumping OAI and Anthropic together. Really, Anthropic are the evil ones. You can make the case OAI are evil too if you want, but Anthropic are very clearly significantly worse and they aren't even in the same b
329.
by nullbio 3mo ago
Weird take. I meant to say GPT, not ChatGPT. Was referring to the model, not the harness.
330.
by nullbio 3mo ago
That's what they're going for, but it's an impossible goal. There is always nuance in decisions being made, and if you can't direct the output on tasks that can have equally correct outcomes, you're just going to en

More ›