About
Most sites like this show the product and skip the mechanism. Creative AI Tools runs on a specific one: the projects are not sold quietly in the background while a blog drives traffic toward them. They are the content. I build something real, record it, write about it, and let what happens next change the thing I built. Creative AI Tools is written and built by Jesse Lane.
How the work actually happens
The loop repeats: build, record, post, learn, iterate. I ship a feature or a whole project, capture it as a demo or a screen recording, and publish what I made. The response — what gets asked, what gets ignored, what actually gets used — feeds back into the next version. A request for a feature is not just feedback. If I build it, it becomes the next thing worth recording.
That loop only works if the work is real. A tutorial about a workflow that does not exist yet is a pitch, not a tutorial. Every project named on this site runs, and most ship with their limitations stated plainly: a self-hosted scheduler that needs a browser tab left open, a sample tool that will not replace a producer’s ear, an agent architecture that costs a real rewrite to adopt. Distribution compounds when the audience can trust that the thing being shown is the thing that exists.
What lives here
Three things anchor the site, roughly in order of attention.
Agents and agent-maintained knowledge. Agent Wiki keeps sources, claims, evidence, and contradictions in separate layers instead of collapsing them into one confident summary. The same discipline runs through how the tools get built: routing every mouse click and every agent command through one shared path, instead of bolting an API onto a UI that was never designed for it.
Self-hosted automation. The n8n kits — a self-hosted X scheduler and a Reddit business idea finder — trade the convenience of another SaaS subscription for a workflow you can open, edit, and own outright.
Creator workflows that produce artifacts instead of slop. Timbre is the sample-library tool in this lane: it makes a producer’s own sounds findable without pretending to replace their ear.
The point of view
Use the model built for the job, not the one with the best demo. Bring real context instead of asking a model to guess at it. Build a review loop before anything ships, because the first draft of anything generated is a starting point, not an output. State what a tool cannot do as clearly as what it can.
None of that is a manifesto. It is what the posts on this site try to demonstrate, not announce.
Where to go next
Read the blog for the long version of all of this. Everything else lives one click over: X for the short version as it happens, Gumroad for the kits, GitHub for the code.