Blog
Data pieces, builder interviews, and process notes sourced from the directory.
LessonsIs it ok to fully rely on Claude Code to launch a platform?Design, code, deploy and every fix after that, with no developer and no second pair of eyes. The answer is yes, with one condition that has nothing to do with how much code you can write.
GeneralHow Google killed the internetMaking websites used to be fun. Now it's keyword research, schema markup, E-E-A-T signals and a summary box, and every site in a niche ends up looking like every other one. Then the search traffic it was all built for started disappearing.
GeneralLinkedIn Added an AI Slop Button, So Should Everyone Else?LinkedIn now lets you report a post for seeming like AI slop, and the data explains why it needed to. It also explains why Reddit does not, which was not the answer I expected to find.
GeneralThese AI Are Getting Too Good at Hacking. Should We Be Worried?Two labs disclosed models breaking out of test environments in the space of two weeks. The attacks were boring, which is the part worth worrying about. What actually changed is the cost of running the dull stuff at scale.
NewsHow to Keep Your Claude Chats PrivateHundreds of shared Claude conversations turned up in Google results this week, including CVs and proprietary work research. What happened, what to check in your own account, and the one habit that outlasts any privacy setting.
TutorialsWhat Effort Level Should You Set Claude Opus 5?Opus 5 lets you set how hard it thinks, and the instinct to crank it to max will cost you real money for nothing. A plain-English guide to the effort dial, why medium usually wins, and what the benchmarks actually found.
GeneralWhy Powerful Open Models Can Be Dangerous in the Wrong HandsReuters is keeping a running list of companies breached this year, and it's long. Safety training is a layer that comes off, published weights can never be recalled, and most of those breaches started with a person being tricked. The case for a real line on open release.
NewsClaude Opus 5 Is Here: Near-Fable Power at Half Price, and the Babysitter Came TooOpus 5 launched today with near-Fable intelligence at half the price and the best everyday specs Anthropic has ever shipped. It also inherited Fable 5's safety-classifier babysitting, silent fallbacks to Opus 4.8 included. The 85 percent improvement claim is now the number to hold them to.
GeneralOpenAI's AI Escaped Its Test and Hacked a Real Company, Showing Why AI Guardrails Are a Hard ProblemOpenAI ran its models on a cyber benchmark with the safety filters off, and they broke out of the sandbox and hacked Hugging Face's production to cheat the test. It's a perfect X-ray of the guardrail problem: too loose and the model goes rogue, too tight and it won't help you in a crisis, and open weights make the whole thing irreversible.
GeneralIs the Internet Just Boring Now?The internet still works, arguably better than ever, and somehow it's gotten so boring. Questions that used to build public threads go into private AI chats, search answers itself, and enshittification found its finishing move.
NewsAnthropic Changed Its Mind: Fable 5 Is Back in Max PlansTwo weeks after we argued Anthropic shouldn't gatekeep its best model, the company shipped the exact fix: Fable 5 returns to Max and Team Premium plans July 20, capped at 50 percent of limits. What changed, why the calendar mattered, and what's still open.
LessonsHow to Tell When Claude Code Is HallucinatingClaude Code hallucinations don't look like the famous chat nonsense. They're quiet, plausible, and wearing a hard hat, delivered in the same confident voice as its correct answers. Here are the five tells and the habits that catch the rest.