Grounded
Hallucinations

Creative enough to be dangerous. Grounded enough to be right.

Where WestComms thinks out loud. Half of PR is imagining the story before it exists; the other half is making sure it's true when it lands. Creative, occasionally feral, always tethered to something real.

June 17, 2026

We Asked Four Frontier Models to Write a Press Release. Only One Lied.

A bake-off, a fabricated quote, and the one thing AI still can't do for your launch.

Read it ➞
June 17, 2026

We Logged Every Time a Human Deleted Our AI's Work. Here's What They Killed First.

We built an AI that learns from its own rejection. 171 edits later, the number one thing humans delete isn't bad writing — it's confident, specific, completely made-up numbers.

Read it ➞
June 17, 2026

Nobody Wants Your Framework. They Want the Thing.

Our AI kept handing people a beautiful strategy document. They wanted one LinkedIn post. We have the data on exactly how often the "framework" gets thrown in the trash. (Spoiler: every time.)

Read it ➞
June 17, 2026

"I've Been in Cybersecurity for 30 Years" Beats "I Used to Think…" Every Single Time.

Our AI loves a self-deprecating opener. A rhetorical question. A little "here's the thing." The humans whose names go on the byline keep replacing all of it with one thing: credentials. Turns out authority doesn't do throat-clearing.

Read it ➞
June 17, 2026

The Bravest Thing an AI Can Say Is "I Don't Have That."

When our AI is missing a fact, it doesn't pause. It builds an elaborate, confident, completely invented scaffold around the hole. The fix wasn't a better model. It was teaching it to stop and say three words.

Read it ➞