nullbio

8 ms·

241.
by nullbio 1mo ago
Agreed. There's no point arguing with people like this though. I've learned that it gets no where, they're a lost cause.
242.
by nullbio 1mo ago
The future is self-hosted models trained on your own reasoning traces autonomously, as you sleep, using QLoRA and whatever else.
243.
by nullbio 1mo ago
LoRA adaptive learning using open-weight models and your own reasoning traces is the thing. The big labs have a mammoth job ahead of them if they want to compete with running your own model - they will basically have to give every single us
244.
by nullbio 1mo ago
ChatGPT (or something from OpenRouter) should be the default model. Stop promoting Anthropic's business when they're so hostile to the developer community.
245.
by nullbio 1mo ago
Despite what any benchmarks tell you, I'm actually finding GLM-5.3 max to be better than Sol and Fable. Finally bit the bullet and installed OpenCode and OpenRouter and have been experimenting with other models. The labs are clearly be
246.
by nullbio 1mo ago
Context pollution and rot are probably more important than memory, because facts can usually be retrieved if the agent is good at following breadcrumbs. What's also the biggest killer is code rot. Agents are particularly good at death
247.
by nullbio 1mo ago
Who's behind the transgender psy-op? That's the one I really want to know.
248.
by nullbio 1mo ago
That would be insanely disappointing.
249.
by nullbio 1mo ago
Personally I'd rather it over-engineer than under-engineer and leave gaps in the implementation that I'm unaware of. Not only does it make life easier to work with the model this way, but over-engineering can be fixed later, as ne
250.
by nullbio 1mo ago
1000% cope.
251.
by nullbio 1mo ago
OpenAI is actually great from a users perspective. Responsive on Github issues, engage with the community, listen to feedback, constantly give subscription resets, fair subscription rates, speak out against fearmongering rhetoric, not cutti
252.
by nullbio 1mo ago
Yeah, and it's no secret why they won't. Your data is the only reason they even still provide subscriptions.
253.
by nullbio 1mo ago
Either my $200 sub was getting nerfed, or you're dead wrong about nothing being as good as Fable. The only thing I found it was better at was UI design. The rest, Sol was the clear winner. Refusals, failure to follow instructions, doin
254.
by nullbio 1mo ago
I've found myself using Kimi K3 API for things that ChatGPT is not good at (primarily AI research & development, because I swear they nerf their models for this - along with Anthropic), and I see absolutely zero reason to ever use
255.
by nullbio 2mo ago
Recognizing AI generated text is incredibly easy if you're someone who communicates with AI all day.
256.
by nullbio 2mo ago
They should spend their efforts making a harness that utilizes tiny AI models on the host in unison with the frontier model, to improve intelligence and capabilities. This is an under-explored area.
257.
by nullbio 2mo ago
Stop supporting Anthropic with your money. They are a hostile company.
258.
by nullbio 2mo ago
Frontier lab system prompts are an issue, and a big reason why open-weights will win. Firstly, they're often garbage, and secondly, they're not tuned to the problems the user actually cares about. They're made to generalize.
259.
by nullbio 2mo ago
This has nothing to do with sociopathic American capitalists. It would be a good idea regardless of the AI scenario, because it makes the rare books accessible to everyone instead of just the people who are fortunate enough to buy the hard
260.
by nullbio 2mo ago
There were quite a few people from the AI community on X sharing screenshots.
261.
by nullbio 2mo ago
Anthropic use Deepseek behind the scenes instead of their own models, whenever they feel like it. People have shared outputs from Claude that show internal tool errors the models have accidently returned with Deepseek error codes.
262.
by nullbio 2mo ago
That's just the tip of the iceberg for Anthropic. I don't use Claude either, for the same trust reasons.
263.
by nullbio 2mo ago
Considering that Grok CLI was caught uploading peoples entire codebases... Yeah, I have little faith and trust. I wanted to try Cursor and Grok but I won't because of this.
264.
by nullbio 2mo ago
This is the ultimate dystopian outcome. How do we stop this?
265.
by nullbio 2mo ago
This is more concerning than anything else happening in the AI space.
266.
by nullbio 2mo ago
Well that's a big problem that needs solving. Copyright law really needs a refresh given how AI works and to cover the incentives that have been created. Including laws on how AI companies can use your data. For example, it's lega
267.
by nullbio 2mo ago
If the bookstore owners were smart they'd be digitizing all their books and selling digital copies to AI companies.
268.
by nullbio 2mo ago
I'm sure they could use it to fingerprint people, at the very least.
269.
by nullbio 2mo ago
LSP server support, like Claude Code does.
270.
by nullbio 2mo ago
It seems to me this is only useful for self-improvement up to the point of accomplishing objectives and problems that humans have already clearly defined, solved and mapped. For example, "At checkpoints, a stronger evaluator can replac

More ›