BK

be the person your AI thinks you are

last week, i built a small GTM stack for myself a few agents watching reddit/x/linkedIn, pulling quotes, and turning market signals into post ideas founder-led GTM breaks when the founder is the system of record @tadata_team turns GTM judgment into lightweight infrastructure

Every security flaw discovered by AI was there before AI, waiting to be discovered either by people or by AI. The world has never been good at securing computer systems; finally with AI we are going to get good.

When I was at Google, I wasn’t afraid to make mistakes in production. The general consensus was: if a well intentioned engineer manages to bring the system down, then we better fix the damn system. The Delve founders should definitely be held accountable if this is true. But this really is bigger than them. They didn’t even try too hard to be sleazy, they just followed the Silicon Valley playbook. 1. Drop out of school as a status symbol, completely missing that correlation is not causation. Dropping out does not make you a genius. 2. Start a business with 0 mission (no 21yo dreams of compliance) 3. Fake it till you make it (hide human labor behind the grandeur of AI features) 4. Raise an obscene amount of money because you can and because those losers who stayed to finish their degrees will be jelly. This is the playbook. The biggest culprits are the ones who made it and uphold it. If you’re not allowed to drink before 21 but are allowed to raise 30m on a compliance idea with no due diligence from investors, then maybe something is really really wrong with the system.

Claude: "Go ahead and paste your API key here when you have it" Me: *pastes API key* Claude: "NO you fucking IDIOT"

why does every ad that shows chess always set up the king and queen on the wrong squares?

The head of US cybersecurity just demonstrated exactly why every company blocks ChatGPT by default. DHS built an internal AI tool called DHSChat specifically designed to keep sensitive data inside federal networks. ChatGPT was blocked department-wide. Gottumukkala requested a special exception anyway. He got it. Then he uploaded documents marked “for official use only” until automated alerts started firing. One official summarized it: “He forced CISA’s hand into making them give him ChatGPT, and then he abused it.” This is the pattern playing out across every enterprise right now. Executives demand access to the shiny consumer tool. IT builds a secure alternative. Leadership bypasses it. Data leaks. Samsung banned ChatGPT after engineers uploaded source code. Amazon warned employees after finding internal data in ChatGPT outputs. Apple, JPMorgan, Verizon, Deutsche Bank all blocked it. The lesson keeps repeating: the people with the most access to sensitive information are the ones most likely to bypass controls, because they assume the rules exist for everyone else. Public ChatGPT sends everything you upload to OpenAI. That data can train future models. It can surface in responses to other users. 800 million people use ChatGPT. Your “for official use only” document is now part of that pool. Enterprise AI policies exist because convenience always beats security until something breaks.

AI lowered the barrier to entry but raised the bar for software storytelling is the new moat

the iphone air is the most beautiful piece of hardware in a long, long time

man what a beautiful ad