I find that one of the hardest things about being self-directed in what I work on is not doing “all the things”. Being relatively new to working in Bitcoin, I see so much opportunity of things to work on and research directions to pursue, like a kid in a candy shop. I’ve spun up a handful of projects in the last few months, none finished, so need to focus on improving my self management to close things out, and move on, where necessary!

One thing I miss about old @npub10uth...xdlq @npub1guh5...6hjy @ODELL is the weekly update on Samourai Whirlpool unspent capacity, and the number climbing each and every week Privacy tools are essential in the face of a global panopticon

I should have done this a bit earlier, but now that it has stopped raining in Perth, and it’s all about Bloody Ripper Days (BRDs), I’ve moved “the office” outside

Wow, on a cursory first look this looks pretty amazing… (and multimodal models are apparently now called “omni” modal) … Qwen3-Omni “capable of understanding text, audio, images, and video, as well as generating speech in real time.” Will be adding this to the self-hosted stack I run at home to put it through its paces!

GitHub

GitHub - QwenLM/Qwen3-Omni: Qwen3-omni is a natively end-to-end, omni-modal LLM developed by the Qwen team at Alibaba Cloud, capable of understanding text, audio, images, and video, as well as generating speech in real time.

Qwen3-omni is a natively end-to-end, omni-modal LLM developed by the Qwen team at Alibaba Cloud, capable of understanding text, audio, images, and ...

Dear Australian Nostriches Where does one get a milk crate from these days?

I get a lot of utility out of Claude and Claude Code but I do want private AI to succeed.. … so I just signed up for a year with @npub10hpc...dxkr 🎉 I love services that are privacy preserving and can be paid for in Bitcoin. Adding it to the stack: - @npub1wlx4...zy3z - @npub1xsgy...zazx - @npub1mea2...qane - Nostr - @npub10hpc...dxkr

How should Bitcoiners react if quantum computers become a reality quicker than expected? The best time to plan is when you have the time to plan! Together with Clara S we put together this playbook. What do you think we got right? What did we miss?

GitHub

GitHub - deadmanoz/bitcoin-crqc-response-playbook

Contribute to deadmanoz/bitcoin-crqc-response-playbook development by creating an account on GitHub.

Would Anthropic purposefully nerf Claude Code so you use up more tokens/usage, hit limits and be pressed to upgrade?

My top 3 biggest frustrations with vibe coding as of September 2025: (1) lacking visibility of sub-agents actions, I’ve had agents purport to refactor code when they did nothing of the sort (2) lacking awareness of temporary fixes, placeholders, “a real implementation would…” (3) loss of context between compactions, so much subtlety that has been built up over the course of a session is lost between compactions What’s yours?

Fast approaching… ‼️ @npub10697...z835 @npub1jss4...fmzy @npub1wpmy...ead6

The velocity of change when vibe coding is intoxicating, but as with most intoxication, there’s often consequences to this velocity🍻

I’ve been burnt enough recently by Claude Code et al going totally off piste and removing functionality, using placeholders etc when on “allow all edits” that I’m only going to use that setting when the changes are very straightforward. So much rework and frustration.. I also had sub-agents totally gaslight me - purport to refactor some code and make absolutely no changes!! Even with manually approving edits, development is still much faster than without these tools

This is to say nothing about the bias of the ABC’s general editorial leaning but I think that some of the kids programming is pretty good Been digging the new Story Salad Quest Mode with the 6 year old. And the “Henrietta the horse” episode in particular

ABC Kids listen

The story of Matt's mighty misadventure - ABC Kids listen

Normally you'd send in just three words for Story Salad, but this time we're hungry for more... Join Matt and his team of expert improvisers as th...

Definitely seem to be hitting limits on Claude Code quicker recently. And Codex really does seem to be a “better engineer”. And larger context is quite useful in some circumstances But I “trust” Anthropic more than I do OpenAI.. First world problems

Developing a small Bitcoin TX classifier using 3 different techniques to experiment with the “best” way to build things: 1) Vanilla Claude Code 2) Claude Code with a custom “mvp-spec-designer” agent 3) OpenAI Codex Initial thoughts, Codex is much better than when I tried it maybe a fortnight ago, OpenAI catching up?