Jev is still so new that some of the best free GitHub repos for it have under 1,000 stars. In this video, I go through seven of them you can plug into Claude Code or Codex today, from voice browsing and computer use to an MCP that gives Claude a set of Jev judgment tools. The seventh is where to find ideas worth shipping.

Jev doesn't work with text at all. It speaks in probabilities, which changes what you can hand off from Claude. In this video, I break down what Jev is, then show it scanning 584 pages for SEO, scoring our outreach leads and running my trading bots, with the costs against Opus and Haiku. Then I set it up in Claude Code with one prompt.

The new Claude model is live! Honestly, very impressed. It's probably the best AI model in the market right now. Tested it for 8 hours, then filmed a video (you can watch below). I also attached a "Project Audit" prompt you can run in your Claude Code to use Opus to audit past projects and make your sessions more efficient.

A week or two after a new model lands, Claude starts getting dumb again. I go through five free GitHub repos that fix it: persistent memory, two ways to stop it burying the answer, a skill that strips the AI tells out of your writing, and one file based on Andrej Karpathy's principles.

My Instagram is at 101,000 followers and does over $15,000 a month, off a contact list of almost 80,000 built with Claude and ManyChat. In this video, I map the whole system: the folders and skills behind my ideation engine, the tools I scrape and schedule with, the funnel that turns a comment into a contact, and where the $15,000 actually comes from.