Claude is brilliant at everything. Except knowing what happened in your week.
Claude is one of the most capable AI models in the world. It can write a LinkedIn post on any topic you give it. The problem is that 'any topic' is the hardest part — and Claude has no way to know what actually happened to you this week. Say Something does.
General intelligence vs. specific excavation.
“Trust isn’t built overnight — it’s earned through consistent actions over time. After years of working with clients, here’s what I’ve learned: 1) Do what you say you’ll do, every time. 2) Be transparent when things go wrong. 3) Put their goals ahead of your metrics. The clients who stick around for years aren’t just buying your service — they’re investing in your reliability. What’s your most important trust-building practice?”
A client called me on Monday to say they were pulling the contract. Not an email — a call. That’s how I knew they were serious.
We’d missed a deliverable by six days. Not catastrophically — but in their business, six days is a quarter-point shift in the earnings call. I didn’t know that. I should have asked.
They stayed. But only because I called back within the hour and asked the question I should have asked at the start of the engagement: “What does a bad six days look like for you specifically?” Every client has a different answer. I’ve started asking it in the first week now.
Claude knows everything except what you lived this week.
The topic problem
Claude needs a prompt. 'Write me a LinkedIn post about X' produces a post about X — competent, readable, and interchangeable with every other post about X. Say Something doesn't start with a topic. It starts with your week and finds the topic inside it.
Voice vs. output
You can ask Claude to 'write in my voice' — but it can only approximate voice from examples, not from understanding your situation, your stakes, or the specific tension in Tuesday's meeting. Say Something surfaces those details through questions.
No kill list
Claude will write whatever you ask, including numbered lists of wisdom, inspirational sign-offs, and engagement bait. Say Something blocks them by default. The 14-rule kill list exists because most AI-assisted LinkedIn posts fail the same way.
Use Claude for everything else. Use Say Something for this.
Claude is genuinely exceptional — one of the best tools we’ve built at Anthropic. For research, writing, analysis, coding, and almost anything else you can describe, it’s hard to beat.
But LinkedIn posts that sound like you aren’t something Claude can do from a prompt. It doesn’t know what happened. Say Something does — because it asks. Try the conversation here and see what five minutes produces.
Common questions.
Can’t I just give Claude a detailed prompt about my week?
You can — and it’ll write a better post than a generic one. But writing a good prompt about your week is itself the hard part. Say Something does the prompting for you, asking the specific questions that pull out what’s actually interesting. The conversation is the product.
Is Say Something built on Claude?
Yes. Say Something uses Claude under the hood. But the system prompt, the interview structure, the kill list, and the post format are all built on top of it — designed specifically for LinkedIn authenticity. Using Claude directly gives you the model; using Say Something gives you the workflow.
What if I give Claude my past posts as examples?
That captures your style — how you tend to write. It doesn’t capture substance — what you actually have to say this week. Style matching is easier than excavating new stories. Say Something does both: it learns your tone and asks the questions that surface the specific thing worth posting about.
Is Say Something free?
Yes. You can write posts, grade existing ones, and check for AI-sounding language — all free, no account required.