Corey McClain

Every AI right now is a "yes man", and I'm soooo tired of it. Maybe you are too. I don't want an LLM that claps for everything I do. I want the one that tells me my writing is weak, my logic falls apart halfway through, and I've been lying to myself about the thing I keep dodging. You can engineer an AI with a spine. Here's the exact setup I use: 1. Write a "critic" skill. A markdown file whose only job is to attack the work. It hunts for the weakest claim, the dodged question, the thing you're avoiding. Every agent that loads it gets a spine by default. 2. Run two agents against each other. One builds, one tears it apart. The builder has to defend or fix. You get the truth from the friction instead of from a single agent trying to please you. 3. Score against evals. Give it a rubric with real criteria and have it rate the work 1-10 on each. A number tied to a standard forces honesty that "what do you think?" will always dance around. 4. Make it cite evidence for every criticism. Set a rule: no critique without a concrete example from the work and a reason it matters. It has to show its work, so it can flatter you or hand wave. 5. Give it a kill criterion. Tell it the exact condition where it has to recommend you stop entirely. "If X isn't true, tell me to shut this down." Most agents will optimize a doomed idea forever because you gave them permission to call it dead. I've learned that the default AI claps for you. That's cool for a bit but it gets old quick. If you actually want to build exceptional products, create top 1% work, and grow as a person, you need an AI that pushes you harder than it flatters you. We're in the age of agency. The people who win are the ones who go build the version that tells them the truth. Best time to build there's ever been. Go get 'em.

maybe i'm going crazy but i really can't read chatgpt outputs anymore. the structure of the response is so schizophrenic.

I've been going on about how ChatGPT Work (and Cowork) are missed opportunities for knowledge workers, and to illustrate that take a look at Google's NotebookLM answering the same question as ChatGPT Work, with the same 70+ files. It centers process & sources, not just outputs.

Evening! We’ve gotten lots of great feedback on the new ChatGPT desktop app (which we didn't get totally quite right on the first try), and as a result, we've made some changes. 1/ ChatGPT conversation history and projects are now visible in the sidebar. Also, your Chat and Work history now sync across web, mobile, and desktop. Local tasks still stay on your computer. 2/ You can now easily switch between Chat and Work modes inside ChatGPT on desktop, which is now also consistent with how it shows on web and mobile. 3/ Nothing is changing for users on Codex mode. It's still the OG and best at what it does. And overall we're continuing to fix paper cuts and improve performance, reliability, and efficiency. Keep up the feedback, hope you like the updates!

Can someone explain me how the new Kimi can be so insanely good? The story has been "chinese labs have much less compute but they still stay close to the frontier through distillation". These results seem impossible to explain through distillation alone.