Thoughts
Raw thoughts.
Stuff I've thought about enough to write down. Most of it comes from community calls, YouTube comments, networking conversations, or chats with business owners.
It's the driver, not the model
Everyone can run the same models, yet the outcomes are wildly different. If a setup you could copy today is getting great results, the bottleneck isn't the model. It's the driver.
Benchmarks shouldn't decide which AI model you use
A new benchmark exposed how gamed the old ones were. Which is why I pick models by feel, not leaderboards.
The 'AI consultant' label has an expiration date
Take the label while the window is open. Just don't confuse it with the job.
The what vs. the how
AI overwhelm is a filter problem. Knowing the 'what' doesn't require learning the 'how.'
The real problem with AI slop
Using AI to write isn't the problem. Skipping the review step is what costs you trust.