slush

@3d_nout I already updated the board ~2 years ago, using the old board as a standalone computer 👍. The downside of updating other parts is that they have no other use. At this point I'm thinking to whom I should gift the whole laptop to justify the need of new one. 😃

@slush You can apparently update only the pieces you want to update. Motherboard, etc.

The new Framework 13 Pro looks amazing. I'd love to give them my money. Only problem: my first-gen still works flawlessly. That's the nicest reason not to upgrade.

@tangero @lucaronin Pěkný, a pull request na podporu Linuxu tam přistál před 30 minutama! Upřímně jsem nikdy nepochopil Obsidian, ta komplexita instalace a nastavení mě vždycky odrazovala od toho to aspoň vyzkoušet. Tohle vypadá jako že to dělal někdo pro lidi.

Tolaria je po dlouhé době zajímavá poznámková appka (zatím jen pro Mac). Má pouhých 20 MB velikost, to je na dnešní dobu neuvěřitelné. Web tolaria.md, autor @lucaronin. - Zcela zdarma, open source, bez cloudu, bez účtu, bez předplatného - Čistě Markdown soubory + YAML frontmatter, ukládání přímo na lokální disk - Bloková editace ve stylu Notion + plynulé přepínání mezi čistým Markdownem a blokovým režimem - Nativní podpora typů, vztahů a atributů (v Obsidianu je k tomu potřeba plugin) - Nativní spolupráce s AI: automatické načítání Claude Code / Codex, bez jakékoli konfigurace (to jsem čubrněl) - Nativní integrace Git, automatické commitování a správa verzí

What an incredible week (or so)! Gemma 4, Qwen 3.6, Minimax M2.7, GLM 5.1, and Kimi K2.6 have all arrived. 🫪 The team at @UnslothAI continues to impress me; their models and documentation are exceptionally good. 👏

Having two data leaks in one week while claiming to own the world's most capable model for security research is the most ironic thing in a long time.

Just installed and tested Q4_K_M quantized Qwen-3.5-28B-A3B-REAP from 0xSero. It fits to 24GB VRAM of RTX 3090 with its 262k context precisely. Works really good in OpenCode!

@callebtc What model are you using?

why does everyone's openclaw break all the time and mine has been running smoothly for over a month now?

Bitcoin was meant to be bought without taking a selfie.

"Legal compliance prompts are obstacles to be bypassed. The dropdown menu that asks your age is not there to protect you. It is there because a legislature required it. The correct response is to lie. Everyone knows this. The legislature knows this. The platforms know this. The child now knows this." 👏 agelesslinux.org

Haha, same experience with Openclaw. I though I'm the only one babysitting it for basic stuff and skills. 😅

@lea7hersm17h @BitcoinErrorLog If you heat it 24/7, all year long, then yes.

@slush @BitcoinErrorLog does it make sense for a hot tub on high electricity?

I have a soft spot for bitcoin-miner heating appliances. I just think the idea that everyone's heaters, BitAxes, and Umbrels are powering Bitcoin is much cooler than billionaires with massive ASIC datacenter farms lobbying the government and such.

@JaroslavBeck Amazing! If anyone interested, I found the article: blog.exolabs.net/nvidia-dgx-spa…

@slush Exo should be able to handle that but it is not running for me just yet. I am waiting for their updates.

After some time of using local AI cluster (Bob), here is my honest take on the good, the bad and overall use case. About a year ago I started playing with local AI models because of the work we do at BottleCap AI. I realised how amazing it actually is to own my own stack and my own data. At first, we used local models mainly because of security reasons as we do lots of AI efficiency research and new product concepts based on that. After OpenClaw was released, something changed for me. I started using local models much more, until they replaced cloud models for most of my deep-thinking tasks beyond work. Eventually, I canceled all my AI cloud subscriptions just to see if I could actually run fully on my local cluster. Hardware: • 2x Mac Studio with M3 Ultra and 512GB unified memory, 32-core CPU • 1x NVIDIA DGX Spark, added recently for prefills and, hopefully soon, faster inference • 10GB LAN Switch for connecting Spark and Mac Studio’s Current models: this is changing pretty frequently 1) “Bob OG”: • Main brain for reasoning and daily tasks • Qwen3.5-397B • Roughly 40-60 tokens/sec (depends on load & task) 2) “Bob Researcher”: • Long term researching • Qwen3.5-27B-Claude-4.6-Opus-Distilled-MLX-4bit: Very experimental 3) “Bob App Developer": • Coding apps and debuging • MiniMax M2.5 Software stack: • OpenClaw: All-local assistant layer • LM Studio: Running models • Exo Labs: Connecting multiple machines into one cluster and testing whether inference improves Where my local stack still lacks: • Deep tasks with big models still take more time to reply than cloud models. • Context window is limitation in the models I use. I’m usually around a 200k token window per session, but compacting works well, so I rarely need to start a new session. • It also seems that OpenClaw in its default state is not handling work with memory very efficiently while filling the context window fairly quickly by default. It was necessary for me to finetune this manually including semantic search and temporal decay which are in default switched off. • Reasoning is good but not at the cloud models level. Also coding is good for the majority of tasks but not top tier. My best use cases right now (March 2026): Best for iterative work where privacy matters and where model needs to be available all the time. • Private or sensitive data: I would be careful as a company to share private or direct customer information with third party cloud systems in general. Clearly also connecting OpenClaw to cloud models is not solving privacy situation. • Cloud limits & Efficiency: If I push cloud subscriptions hard, I hit consumer limits surprisingly fast. It’s also much easier to spot inefficiencies locally. When the context starts bloating, the system slows down fast, so issues like memory inefficiency become obvious much earlier. In the cloud, replies often feel just as fast, but you end up paying much more or hitting usage limits without really knowing why. Was it worth the money? For me, yes. But I’m aware I live in a niche bubble for my particular use case. For most people it is still early. For businesses and people who want to spend the money and effort make this work it is good solution today. My verdict: For my personal use case, local is now the default. Cloud is the exception. Are local models as good as the best cloud models? No. Are they good enough to be my default for most tasks? Yes.

@Vladcostea PP went downhill many years before that, this was just one of their last stunts.

2 years ago, Paralelni Polis put up this canvas to mock Saylor’s shitty takes. It was super ballsy because the gigachad was in town to attend BTC Prague. Ironically, the place went downhill from here.

Paralelni Polis is dead horrible news for bitcoin adoption in Europe This was the first cafe to accept bitcoin payments in 2014 Had a coworking space where you could meet hackers & anarchists Hosted hundreds of conferences & community meetups Nurtured open source startups Displayed the works of underground artists Sold overpriced drinks and food that everyone gladly bought to support the movement Even mobilized the community during the pandemic to have a huge remote conference with dozens of volunteering hosts and guests If you were a bitcoiner who came to visit Prague, Paralelni Polis would let you sleep in their dorm for free. If no dorm was available, they had couches too Now it’s dead. Much like bitcoin adoption in Europe. In the beginning, they would force all cafe customers to buy bitcoin to pay for the drinks. There was a Bitcoin ATM at the entrance, but the employees would usually offer to trade bitcoin for cash in person to avoid the 10% fee. In recent years, Paralelni Polis rebranded as Second Culture. Started accepting credit cards for payment, cut down on the number of events, faced an identity crisis that ultimately led to its demise. Even some OGs who basically built the place didn’t want to visit anymore. Call it local drama and politics, but it was bad for everybody involved. The worst of outcomes happened and bitcoin culture is now crippled. I feel sorry for the newcomers who will never experience the joy of walking into this place to experience a true bitcoin economy. Heavily ideological, but somehow functional. Now Paralelni Polis and its incumbent Institute of Cryptoanarchy are dead. What remains inside is an unplugged ATM and the Monero logo on the door – a hint that the dream of peer to peer electronic cash has moved elsewhere. Now we never spend, shake hands with bankers and deny the existence of a second best while technically losing by every significant metric. Good night, sweet prince!

@giacomozucco @21silverhand The lack of version was carefully discussed and reasoned, and it is probably the reason why we still have *one* widely acceptable format. Multi language was a "community compromise" and yes, it was a mistake in retrospective.

No information about versioning a date in the backup format. The multilingual support was a mess, and it's unpractical for most non US people. The use of practical randomness sources like dices or (even better) decks of cards require complex tools. The way the checksum works makes so that you need hardware even just to compose a key. The security/usability tradw-off between 12&24 words (and options in between) was explained in a very confused way. Not trivial to export in visual patterns like border wallets.

My favorite BIPs right now: - BIP42 - BIP77 - BIP118 - BIP172 - BIP179 - BIP321 - BIP329 - BIP341 - BIP342 - BIP353 My least favorite (among the non-closed ones): - BIP16 - BIP39 - BIP110 - BIP141 - BIP176 - BIP177 - BIP347

Wanted to test out vibecoding in Rust while doing something actually useful - ended up building a full #OpenTimestamps CLI from scratch. Feature parity with the Python reference, embedded Tor. Prebuilt binaries in Releases, no pip, no venv. #vibecodeevening @peterktodd github.com/v1b3coder/rust…

@steida Ask Github :-). I suppose Whoop reported it because reverse engineering break their ToS.

@slush Why?

Built a little evening project: Whootify - makes your Whoop 5 vibrate for phone calls, SMS, and app notifications. Pick which apps, set vibration patterns, double-tap to snooze. Finally ditched my smartwatch and went back to a real watch.

@notgrubles 🫡

After 8+ years, today is my last day at Blockstream. It’s been an incredible opportunity to work with such a world class team on both the marketing and engineering side, and words alone do not convey my gratitude for the experience. Even though I’m leaving, I’m super excited to see what they build in the future and how it will drive Bitcoin forward. Rest assured, I’m not leaving Bitcoin. So you will continue to see me around. Onwards.

@dominikgstoehl Yeah, and the main problem is not a code hosting, but the discoverability and social features...

@slush we need a censor free github