You can build a whole workflow without dragging a single block. Describe the system you want: an agent that turns an idea into a finished post, with a draft, a hero image, a slide deck, and a voiceover, and Sim lays out every block, wired and parameterized, in front of you.
What you will learn
One prompt, a whole system
A single described intent becomes a real workflow: a parallel candidates container, generation steps, an evaluator, and a scores table, assembled and wired.
Test and fix the workflow
Follow the example run: the image step fails on a missing aspect ratio, Chat updates the parameter, and the next run succeeds.
You inspect before you trust
It shows you the run record, every step, every output, so you inspect before you trust. Then you hand it a real idea and watch candidates get drafted, scored, and pruned.
Deploy from the conversation
One more message and the workflow is live behind an endpoint, from blank canvas to working system without leaving Chat.
From description to blocks
Sim turns your description into real structure: a parallel container drafting candidate posts, generation blocks for the image, the deck, and the voiceover, an evaluator scoring the results, and a table keeping every score. Nothing about the result is special, it's a normal workflow you could have built by hand, which means you can open it, read it, and change it.
Here is the shape of the workflow Sim builds in the video, candidates drafted in parallel, an evaluator scoring them, and the scores kept in a table:
Test and inspect the workflow
In the video, the first test fails because the image block is missing an aspect ratio. Sim reads the error, updates that parameter, and runs the workflow again. Ask Chat to test your workflow, then inspect the logs and outputs before deploying it.
Inspecting the run
The run record shows every step with its timing and output: the draft, the image, the deck, the audio. When you test it on a real idea, the candidates fan out in parallel, the evaluator scores each one, and the table records what was promoted and what was pruned.
Deploying
After testing the workflow, ask Chat to deploy it as an API endpoint. Review the deployment settings and test a request against the endpoint. The workflow remains editable in your workspace.