gm nostriches 🤔

AI has worsened the problem of armchair experts. The only people that benefit from you publishing your unverified and crappy slop of any form, whether it be text, code, images or video, are the AI companies squeezing you for $$$ Please stop polluting the internet with garbage that is not even worth the fraction of a cent worth of disk space it occupies. View quoted note →

It appears now on Twitter, a random idiot with an AI tool has gained traction by writing an article longer than STM32 reference manuals about a BIP-110 vulnerability. However, the vulnerability was completely hallucinated (this post is not in favor of/against BIP-110)

Instead of more clients/features/infra/whatever, there needs to be more things developed long-term and focused on stability. The state of Nostr usability has barely improved pre and post AI slop, if anything it got worse.

Good morning

I will not be running BIP-110 or a similar softfork. Such a fork solves nothing. However, the behavior of Bitcoin Core is also harmful to Bitcoin. The OP_RETURN change had no technical basis and had other problems: - People will not pay 4x more to use OP_RETURN. The “cheaper at small sizes” argument was fixed by Taproot annex already. - SegWit data can be fully pruned (or even not downloaded) while being able to verify supply. OP_RETURN can’t. - Both SegWit and OP_RETURN do not inherently pollute the UTXO set. It is due to the specific implementation of Ordinals.

You were right—I delved deeper, and closed the seams of the final implementation pass. The load-bearing features are gated by a bounded narrow health check with a canonical data plane to ensure the load-bearing product surfaces fit the shape even with chaos goblins

Why do people hand over full access to their computer to an LLM? The vendors are logging everything and these models are unpredictable. Your information will be leaked or used one way or the other.

gm / ga nostriches The thing about vibe-slop businesses is that they will inherently fail due to commoditization, unless you have a differentiating factor. (assuming OpenAI/Anthropic/Google will not steal it)

gm

It seems that part of GPT 5.6’s token efficiency came from reducing the required “thinking budget” for high/xhigh. So, xhigh is no longer truly xhigh.