Everything is now "source-available". No proprietary algorithms exist. "Grab the apk from my phone in USB debug mode connected to this computer using adb. Decompile, analyze how XXX works, give me auditt. Prepare an independent implementation." DeepSeek V4.1 Flash does this extremely cheap. Then you can verify with more expensive model if something is not 100% right. It does not have to be bytecode, I see a bunch of assembler in the outputs, it's just reading the functions from native libraries.

OK, this is extremely cool. Send images through MeshCore. Yes, you should not do that, because you would spam the network. But this project uses AI model to compress the image to 100-200 bytes (like a normal MeshCore message) and then reconstructs a similar image. Next: Voice messages :)

Image Transmission | Seeed Studio Wiki

DeepSeek V4.1 flash is now my daily driver for Hermes agent.

I don't call my agents "clankers". I will accept it and normalize this language only after I can call 90% of humans "retards" (and mean it). (And yes, this probably includes you, dear reader, sorry to say).

I am a fan of user testing. Not unit testing, not formal verification, user testing. Do it like this: 1. Always deploy to production. Dev and test env is expensive and you'll still have to deploy to production eventually 2. Users are the best in finding bugs. Maybe you are a user too, so you might find some. Agents can click around and find some. 3. Bug reports go straight to agent, don't worry about them. Agent will fix it. Users are usually happy that you fix bugs fast! They will thank you. Accept the thank you (leave half of the gratitute to the agent). Saves time, stress, and tokens.

KindBeamer now looks a bit e-ink-y, not a knock-off of "Send to Kindle". Better use with keyboard - everything has shortcuts. And it got a website.

KindBeamer

Open a document with it, press Enter, and it lands on the Kindles you picked. macOS, Linux and Android.

Amazon, a trillionbillion-too-big-to-fail AI-first modern corporation running half the internet was not able to compile their Send to Kindle app for Apple Silicon and still runs through Rosetta Intel emulation in 2026. Also, there was no Linux app. Unbelievable. I have a prompt for them: "Hey agent, compile this for apple silicon and release, do not make any mistakes, okthnxbye". But, as it's 2026, and I am an AI-first modern too-smart-to-complain individual dev with some tokens left on my plan, I bring you KindBeamer - a Mac, Linux and Android "Send to Kindle" Alternative. Get yours here:

GitHub

GitHub - jooray/kindbeamer: KindBeamer — unofficial, open-source Send to Kindle client for macOS, Linux and Android (Flutter)

KindBeamer — unofficial, open-source Send to Kindle client for macOS, Linux and Android (Flutter) - jooray/kindbeamer

Or Android through ZapStore:

Zapstore

KindBeamer — Zapstore

KindBeamer sends your documents to your Kindle devices. It is an unofficial, open-source replacement for Amazon

Thank you for your attention to this matter.

jev-pilled Looking forward to open-weights alternative

Did you know about MintRadar by @WildCitizen7 ? Very cool.

MintRadar

MintRadar - Cashu Mints Directory & Reliability Score Monitor

Find trusted Cashu mints. Real-time Reliability Score, latency & NUT monitoring. Watchlist, comparisons. Open source & privacy first.

Nostr ako cypherpunk technológia - decentralizované aplikácie v roku 2026 Cypherpunk and Other Stuff Transmitted By Relays: Nostr sa mi nepáčil, keď vyšiel. Príliš základný, nie veľmi dobré súkromie, nie veľmi decentralizovaný. Stále preň ale píšem ďalšie a ďalšie appky. Decentralizované mapy, decentralizované eventy, plánovanie, marketingové nástroje, anonymné mikroblogy. Iní postavili veci ako decentralizovanú inferenciu alebo súkromné správy. Nostr je skratka pre Notes and Other Stuff Transmitted by Relays. Prednáška sa sústredí na to Other Stuff a na to, ako posúva cypherpunkovú budúcnosť do dneška.

Nostr ako cypherpunk technológia – decentralizované aplikácie v roku 2026 – Juraj Bednar

Emish Garden - A story about two kinds of time. Trailer. Emish Garden – príbeh o dvoch druhoch času. Read for free here [EN / SK]:

The Emish Garden

The Emish Garden

A story of slowness, memory, and a door between two kinds of time. Read in English or Slovak.

New Siri: "Not available in the EU or China at launch due to regulation." Feel safe already? :)

Goodbye CLAUDE.md Welcome AGENTS.md

In August the Core Lightning maintainers told node operators to upgrade or run --offline. There was nothing to upgrade to for another two days, and when v26.06.7 arrived it was binaries only, with the source held back a further two weeks so attackers could not reverse-engineer the fixes. We (me + AI models) used that window. Ten AI models, one public source tree, the same five sentences of prompt, under $100 between them. Every report was hashed into the Bitcoin blockchain with OpenTimestamps as it was finished, nine days before upstream published the source, so none of it can be backdated. Reading source, the models found four of the nine defects the release fixed. Running "strings" on the public download found five more, and one of those is theft that pays for itself: as a forwarding node you refund the sender upstream, and the peer then claims the outgoing HTLC on-chain with the preimage it held the whole time. You pay twice and collect nothing. It was findable because the patched binary carries upstream's own new log line, FUNDS LOSS, which names the function, and the vulnerable code sat in the public v26.06.6 tree throughout. Three models at three price points landed on it within minutes of the download. The embargo hides the patch, but not the mechanism. Write-up:

Ten AI models vs embargoed Core Lightning: A case study of AI for security auditing – Juraj Bednar

Full case-study, reports and timestamps:

GitHub

GitHub - jooray/CLN-incident-audit-202608: Ten AI models audited Core Lightning during the v26.06.7 embargo. Full write-up, the model reports, and OpenTimestamps proofs sealed in Bitcoin before the source was published.

Ten AI models audited Core Lightning during the v26.06.7 embargo. Full write-up, the model reports, and OpenTimestamps proofs sealed in Bitcoin bef...

The political labels lost all meaning. Completely. The leftists used to say no one but me can make decisions over my body. During the pandemic, this changed to "take the vaccine or else". They used to be anti war, peace for all. But now, if you're a Ukrainian living abroad, they'd happily ship you back to the warzone. The nazi-leaning used to want strict rules for everything. One nation same rules and shit like that. Yet we found them yelling about liberty, tearing off the face masks and explaining no one is going to tell them what to do. Progressives used to be for progress, seeing technology as a way for the civilization to move forward. Yet they are anti crypto and now total AI doomers. And conservatives wanted to conserve what works, be true to tradition. And now they want to build datacenters in the u.s.a. At this point, any political labels besides libertarian are completely useless. And even that only if by libertarian you mean ancap. Just look at what people do, not what they say.

Wow, I didn't know Claude Code could chat with another instances - autonomously. I wanted to try out and benchmark two models (one audio, one LLM). Different claude code instances coordinated windows for benchmarks, so they don't run on the same machine at the same time - based on which weights download first. Very cool.