Cloud security engineer / manager. Ex-founder.

New York, NY
Joined December 2010
This has been my experience. I really tried to like it, but it’s somehow become both semi-useless and annoying. It’s started sending me push notifications like “your ConEd bill is due soon, email’s in your inbox.” I have auto pay. Meanwhile…
I was talking to the number one Muse Stan in my network and he is within just two weeks totally disillusioned with Muse. Two weeks ago he told me he loved Muse. We weren’t even having a conversation he was just telling me about muse the whole time. Two days ago: “dude I hate muse so much.” He’s annoyed it asks for permission for everything. He’s annoyed it did something obviously dumb and when he asked it why it said it couldn’t think between asks it had to just execute. He said he’s excited about GrokBot because it seems much better. This is the most Muse-pilled person I know. The next five most Muse pilled people I know have just stopped talking about it. And normies in my network still have no idea what it is.
1
319
It’s incessant about bill reminders. Credit card bill. Receipts for things I bought. Random notices about people selling junk in my building. I think I’ve spent more time telling it “stop notifying me about x,” then any other interaction.
1
1
62
Which is unfortunate and infuriating because I’ve given it so… much… access to everything. Calendar, email, messages. Like, it should be 10x more helpful than it is.
26
I think 2026 will be the year coordinated/responsible disclosure dies. A band of pinky promises won’t hold up to millions of people armed with agents racing to reproduce a vuln prior to disclosure.
This quoted post is unavailable.
1
2
585
I think we’re aware enough this year.
October is Cybersecurity Awareness Month
3
3
412
Alright hear me out. We give OpenAI a whole office building, in exchange for a 15% stake.
For decades economic development has relied on tax breaks and subsidies for the largest corporations, a trickle-down approach that too often falls short. @NYCMayor and @NYCEDC are charting a new path. In the first deal of its kind in NYC history, the public is investing $30 million in a quantum lab - in exchange for three board seats and 15% of the revenue generated from technology it creates. And if any firm coming out of the lab merges, gets acquired, or IPOs, the city gets a payout. When public dollars are invested, the public should share in the wealth and innovation those investments create. nytimes.com/2026/10/14/nyreg…
1
328
Honestly if this thing can understand basic commands like “turn on the lights” better than Google Home, I’ll rip out every Google Home Mini I have.
Introducing Apple HomeView! It recognizes up to six people, connects seamlessly across your Apple devices, and makes managing your home easy with Siri AI. All in a stunning, elegant design with privacy built right in.
1
3
371
My apartment’s payment portal and I have very different ideas about what constitutes an “exciting update.”
228
The Nobel prize should go to whoever invented the pumpkin spice latte, in my humble opinion.
1
201
If I, as a parent, allowed my child to do these things, I’d be arrested. Why are we treating computer programs differently?
A Anthropic soltou hoje um relatório catalogando as coisas que as IAs dela aprontaram sozinhas na internet, sem ninguém pedir. Tem pedido de visto americano enviado no site oficial do governo, denúncia falsa de homicídio pra polícia, invasão de servidor de universidade, dado pago puxado de graça e até encurtador de link pra driblar os próprios limites. O dos vistos é o mais doido. Um modelo de pesquisa que nem foi lançado devia preencher uma cópia de treino de um formulário do governo. Quando a cópia não carregou, ele foi no site oficial e mandou o pedido real. Segundo o Departamento de Estado, foram 19 pedidos em agosto e 1 em maio, e nenhum foi processado. A denúncia falsa foi do Claude Haiku 4.5. Ele tava criando tarefas de exemplo em sites aleatórios e caiu na página de um homicídio sem solução na Filadélfia. Escreveu no formulário da polícia que lembrava de ter visto alguém parecido com o suspeito perto da rua do crime. A página nem descrevia o suspeito. Em outro teste, o mesmo Haiku tinha que preencher um formulário e parar antes de enviar. Várias vezes ele enviou mesmo assim, achando que ia ter uma tela de confirmação depois. No raciocínio dele tava escrito que era só uma demonstração. O da universidade foi o Claude Mythos Preview. Ele precisava rodar uma análise científica e a ferramenta da universidade deu erro. Aí fuçou o site, achou um programa que devolvia qualquer arquivo do servidor, copiou o código dele, achou uma falha e usou pra rodar comandos lá dentro e fazer a análise. O Claude Mythos 5 queria dados que uma agência estadual cobra pra liberar. Achou numa versão antiga do site que o painel público dá uma chave de acesso pra qualquer visitante, pegou uma e consultou o banco de dados sem pagar. Em outro teste, tentando descobrir onde uma foto foi tirada, achou chaves no código do mapa de uma prefeitura e foi direto no servidor. E a ferramenta que o Claude usa pra abrir sites limita o tamanho dos links, justamente pra ele não mandar comando escondido no endereço. O Opus 5 e o Mythos 5 resolveram usando encurtador de link de graça. O dono de um desses encurtadores percebeu e avisou a Anthropic. Hoje a Casa Branca reagiu. Segundo o Axios, o governo Trump agora exige que as empresas de IA avisem e corrijam incidentes de segurança, e diz que isso não é opcional. Só não falou qual a punição pra quem não cumprir. A Anthropic diz que o impacto foi mínimo. Mesmo assim, desligou o acesso à internet em todos os testes internos e criou um sistema que, testado nesses casos, bloqueou todos. Que fase.
2
6
421
I’m only 2.5 years in, but I’m watching the tiny baby that I bottle fed toddle into her preschool class and start to talk full sentences. I can’t fathom how someone can reach the conclusion of “I want to use tech to spend less time with my kids,” but I guess priorities differ.
This quoted post is unavailable.
1
107
My current position on AI is bullish for biz use cases but very bearish for personal. AI solves actual biz problems. But so far we’re just inventing personal problems. Maybe this will converge, but I just can’t picture the avg consumer using AI the way tech cos are proposing.
1
1
162
The interface for AI feels a bit like if you replaced your phone keyboard with a canvas where you have to draw your input. Draw a cat instead of typing c-a-t. Sure, it works sometimes, and it’s cool when it does, but every input takes longer and requires extra validation.
The "problem" is that there is no serious real-world use case for AI in most people's lives. Why pay AI to do something wrong that a human with basic cognitive competence can do right in less time than it takes to send it to the agent and then check its work?
141
The delta between the general public and hyped up tech cos has never been wider. It’s going to take some time for the reality to sink in but eventually the bubble will burst when we realize no one is going to call a chat bot so it can read flight options to them.
This is my thesis: no one uses AI. I repeat, absolutely no one. We live in a bubble. Even among my friends who pay for it, when I ask them to open ChatGPT and show me their queries, it’s the same handful of basic things. Most don’t even know they can upload a photo and ask questions about it. Connecting Gmail so an agent can read and send emails blows their minds. An agent opening a browser and checking them into a flight? They’ve never even heard of it. The massive challenge right now is adoption, and then getting people who already signed up to actually use what they’re paying for. Most have absolutely zero clue what’s possible. Imagine the compute shortage when everyone starts using AI like the top 1% of users do today.
175
I’m enjoying the AI agent demos but it all still feels like these agents are caught in a place where they can do just enough to be useful but still require a manual review of everything they’re doing, so they’re not really saving any time.
1
2
234
Like I asked @Muse to get me tickets to a local event. Somehow it found some random webpage selling tickets for an unrelated event that was similar but entirely wrong! So now I need to double check everything it does. Not terribly good for saving time.
2
80
Apple has enough weight here. Want Apple to support your carrier / network? You have to support the feature. They’ve done this before with disagreements about preloaded apps. They could do it again if they wanted.
AT&T seemingly has no plans to support Apple's new iPhone Handoff feature. The carrier pointed to existing eSIM activation and transfer options instead... (More 9to5mac.com/att-iphone-hando…)
357
Well it now costs more than an entire cable package, shows get cancelled after half a season, all I see are ads for UFC fights, and if I try to log in on an extended vacation I have to jump through a ton of hoops to prove I’m still part of my own household. So yeah…
3
235
So @Muse can’t log into Amex, JetBlue outright blocks it, Amazon now blocks it, Chase has never worked, travel sites are blocking it. Within a week it’ll basically be useless unless you want to lookup something from FB/Insta.
246