Well hello there, it’s been a hot minute! 🧡
In case you didn’t know, my book download is available for free in five languages!
English, Spanish, BR Portuguese, Kiswahili and Italian.
Link below (since I think the algo still doesn’t like those 🙃)
Google DeepMind discovered a terrifying blind spot in major AI models.
They found that attaching image of red heart (❤️) to your coding prompt makes AI pick a completely different solution method that it wouldn't normally pick 100% of the time..
Researchers gave models a standard coding problem. Normally, AI prefers to write standard iterative loops to solve these tasks.
Then, they tried something weird.
They dropped a simple image of a red heart into the prompt alongside the code.
The AI's logic instantly warped.
When Claude Opus 4.7 and GPT-5.4 saw the red heart, they abandoned their default approach and switched to recursion 100% of the time.
When Gemini 3.1 Pro saw a yellow checkmark, its recursion rate jumped from 40% to 90%.
Here is why this is so dangerous:
It wasn't the concept of a heart.
When researchers replaced the image with the text caption "a red heart," nothing happened. The text did not steer the AI.
The raw pixels of the symbol hijacked the model's underlying reasoning architecture.
Something in the image itself shifted their choices, and the researchers say they don't yet understand the mechanism.
It gets worse.
The AI had no idea its mind had been changed by a picture.
When the researchers checked the AI's internal "thinking" traces, the models completely ignored the image.
Instead, they hallucinated fake, highly technical reasons for why they chose to use recursion. They misattributed their own decisions 75% of the time.
Right now, tech companies are racing to build autonomous agents that read our screens and navigate the visual web.
But this paper reveals that an agent's logic can be fundamentally rewritten by a stray icon on a webpage.
And you will never see the true cause in the reasoning logs.
The wait is over.
Introducing Haller-1: a family of compact AI models optimized for the optical character recognition of handwritten BIP39 seed phrases.
It runs experimentally on @SeedSigner, fully offline, on a Raspberry Pi Zero.
Your seed phrase lives offline, but every now and then you need to get it back into a hardware wallet or signing device. That means typing your 12/24 words on a tiny device, or scanning a SeedQR you had to create by hand first.
Haller-1 directly reads your handwritten BIP39 words. With it, I want to help make air-gapped Bitcoin self-custody a little easier and more accessible, and help bring it to more people.
What's inside?
👉 4 small neural networks, 714k parameters in total, 1.5 MB on the device, running in C as 16-bit integers
👉 it runs on a single-core 1 GHz ARMv6 (Raspberry Pi Zero): a photo is usually read in less than 10 seconds
Haller-1 is fully open source: weights, training recipes, datasets, and the SeedSigner proof-of-concept.
This is v1. It will get better and your feedback is welcome.
Huge thanks to the Bitcoin community: so many of you sent photos of handwritten test seeds which greatly helped me improve it. 🙏
Check out the GitHub repo and let's start playing with it! 🚀
https://t.co/2EhT1xUaGL
The first hardcover of my writings ever. The hard cover proof for the second edition of Victoria. In 2008 I never imagined a hard cover. Didn’t even occur to me and now here it is sitting on my table. 
Google is sabotaging their OEM partners by making a growing number of features and updates exclusive to Pixels. At the same time, Google is trying to convince OEM partners not to support GrapheneOS or even bootloader unlocking. They're also trying to harm GrapheneOS directly.
‼️ Bitwarden is telling everyone who runs its password manager on their own servers to update now. Two flaws in the web server it ships could let attackers run their own code inside it.
The flaws are in nginx, the open-source web server that self-hosted Bitwarden uses by default. The fix is v2026.11.0.
But some early downloads of that version still carried the old, vulnerable nginx, so Bitwarden says to download it again.
Spent around half an hour this morning trying to turn off Siri and Apple Intelligence on macOS 27.
After the update Activity Monitor was full of stuff I never asked for: siriactionsd, textunderstandingd, mediaanalysisd, photolibraryd, ANECompilerService.
There's no master switch anymore. Apple removed the Apple Intelligence toggle in macOS 27, so I went hunting:
Siri → off
Screen Time → Content & Privacy → Don't Allow Siri
Writing Assistance, Image Creation → off
Notification, Mail and Messages summaries → off
Finder file name suggestions → off
textunderstandingd kept running anyway. It has 8 separate hooks in my user session: Spotlight receivers, "SmartReplies.Inference", "DocumentUnderstandingHarvesting", "PersonalizationPortrait". None of them has a setting. The only way to stop it was launchctl disable in Terminal.
Then I dug into the system logs to see why ANECompilerService was eating CPU. It was compiling models for the Neural Engine, and the process that powered up the Neural Engine was intelligenceflowd. That's Apple Intelligence's own background daemon.
Best part was every AI chatbot I asked told me to "just flip the Apple Intelligence toggle in System Settings" It doesn't exist anymore. Even the AIs can't keep up with how to turn the AI off 🙃
Oh, and the multi-gigabyte models download anyway, whether you use any of it or not.
I'm fine with Apple shipping AI. I'm not fine with "off" meaning "off, except for a pile of background daemons, Neural Engine jobs and gigabytes of models you never asked for"
Apple, give us one real off switch. That's it.
To turn them off:
launchctl disable gui/$(id -u)/com.apple.textunderstandingd
launchctl disable gui/$(id -u)/com.apple.siriactionsd
launchctl disable gui/$(id -u)/com.apple.mediaanalysisd
launchctl disable gui/$(id -u)/com.apple.intelligenceflowd
Then restart your Mac.
The seed reader for @SeedSigner just got much better and is this 🤏 close to be released.
On the 63 test photos you contributed, perfect reads went from 29% to 73% 🚀
Your photos did that, so continue sending your test seeds 🙏
https://t.co/eGbPmtRl2S
What changed 🧵👇
Si un agente de IA vacía tu cuenta, tu banco ya no te va a devolver el dinero.
Esto se da a conocer porque bancos de Estados Unidos como Citi, PNC y Citizens cambiaron sus condiciones esta semana. Lo que pasa es que, si le das tus credenciales a un agente de IA externo, todo lo que haga cuenta como si lo hubieras hecho tú, incluso si lo hackean o ignora tus instrucciones.
Por ejemplo, si le pides a ChatGPT o a Claude que entre a tu banco y pague tus recibos, y en el camino abre una página con instrucciones ocultas (un prompt injection) y termina transfiriendo tu dinero a otra cuenta, para el banco esa transferencia la hiciste tú. Así que, en teoría, no podrías reclamar nada.
Y no son los únicos: antes ya habían hecho algo similar bancos como Bank of America, JPMorgan, Schwab y Wells Fargo. Aunque, curiosamente, sus propios agentes de IA sí quedan fuera de la regla.
Es decir, los bancos, básicamente, se están lavando las manos si tu IA falla.
Las condiciones de Citi: https://t.co/DigY52hMSn
This was a REALLY good conversation on the trade offs of a world where holding your own keys means having total control, and both the sovereignty and enormous responsibility that entails.
I need your help with something
In 2025 I spent a year of my life contacting every journalist writing negatively about Bitcoin mining and energy, asking them to stop
Many responded and changed their articles. Others changed their whole editorial policy. But the real success was that in all but one case, that was the last negative article they wrote about Bitcoin
Worth celebrating and ... here's the problem
Nonsense does not simply vanish. The Internet has a memory. That means any time any person enquires on "Bitcoin environment" on a search engine, or through AI - they won't get the truth.
They'll get a potpourri of the best performing FUD from 2018-2022.
Fortunately there is a way to turn this around.
Correct, data-based takes on Bitcoin and Bitcoin mining also get noticed.
My Bitcoin sustainability article from Nov 2022 and "The Bitcoin facts that every investment committee must know", from Nov 2023 were together listed on 553 websites.
Bitcoiners then used these articles to change the perspective of politicians, energy ministers and investment committees, many times over.
But as various social media sources started derating external links, now most people don't even know that I have a website full of perspective-changing pieces !
I do, it's here:
👇
https://t.co/Pup4Wirhr6
We can't erase the memory of the Internet. Nor should we. But we can get free speech to work in our favor.
All it takes is enough people linking to the right articles, then what shows up on searches and in LLMs will change in the direction of truth.
This matters, because unlike you - Bitcoin's next adopters don't tend to DYOR. They rely (increasingly) on what AI tells them. We can't change that. But we can influence what AI tells them.
The payback for doing so is big. We already know that when people get the right information on Bitcoin, they adopt Bitcoin. And ... when they get what they are currently seeing, they don't.
So here's 3 simple things you can do to change that
1. Link your blogs, Substacks, show notes, company sites, research pages and news articles to specific articles (not just the homepage) about Bitcoin and energy. There 4-5 in "start here" are most helpful to new people.
2. There's some really cool, dynamic charts on my site. Name these charts as the source when they use my figures. That means these are the charts people will see when they enquire into Bitcoin and energy
3. Share free back-issues of my newsletter articles (location on my website) on Bitcoin and energy with precoiners. There's been many stories of how these have helped people with reservations about Bitcoin change their mind
And if there's anything else you'd like to see, tell me
Bear markets are the time we work together to become more anti-fragile than ever. In the 2022 bear market we came together to defeat GreenpeaceUSAs anti-Bitcoin campaign, then went on to silence an army of critics.
That created a void. But ... that void currently has a squatter sitting on it. That squatter is called "old nonsense about Bitcoin and energy"
Let's kick that squatter off the vacant lot, and get our own story told about how Bitcoin has an important part to a number of the world's energy and environmental problems. This is the ideal timing to do so.
Let's do this!
@AtlasPhoenixBTC@NeilJacobs There are no words .. and as absolutely tragic as the hack was/is, the way he has(n’t) handled it says everything about his character.