LIVE: The greatest Claude Code workflow ever

200K subscribers, second-channel freedom, and live-stream chaos

Alex opens in full Alex Finn mode: this is supposed to be the “greatest Claude Code workflow ever,” but first he’s celebrating 200K subscribers and thanking the audience for getting him there. He talks about how it took a year and a half to hit 100K, then just two months to hit 200K, and uses that to segue into a lesson about content creation: start a second channel where there’s zero algorithm pressure and you can just post whatever you want.

A detour into motivation, depression, and why action matters

Before touching the tooling, he pauses to respond to a viewer going through a rough patch, and it becomes one of the more human parts of the stream. His advice is to do one productive thing today — one action item — because action creates a chain reaction and helps “reboot” your dopamine, a point he ties back to his own story of being $200,000 in debt for 11 years after college and not really feeling financially stable until 31.

Why he finally embraced Slack after hating it

He jokes for a while about skincare, “grift,” Twitch, and whether his Twitch page looks like a normal person’s Twitch, but eventually lands on the real setup: he used to hate Slack, but a lot of dev tooling is built around it, so he gave it another shot. The reason is practical, not ideological — once you start building a company and hiring people, the loose solo workflow stops being enough.

The actual stack: Linear, Claude Code, Codeex, GitHub, Vercel, Slack

Once he finally gets to the title, he lays out the system he’s using to build Henry Intelligent Machines: Linear, Slack, Vercel, GitHub, Claude Code, and Codeex. He says all of them have free tiers, and the key move is connecting Linear into Claude Code and Codeex so the agent can create projects and issues for whatever feature or app you’re working on.

Linear for planning, GitHub branches for control

This is the most concrete part of the stream: ask the agent to build out the high-level projects and low-level issues in Linear, then tie Linear to GitHub so each issue gets its own branch automatically. Alex’s point is that this keeps everything reviewable and sane — each branch can be tested, reviewed, and merged back into main only when it passes, instead of letting your AI agent freestyle directly into production.

Slack as the team nerve center

Then he adds Slack to the loop: Linear status changes, GitHub pull requests, commits, and merges all get posted into dedicated channels automatically. Even if you’re a one-person shop, he says this creates a clean paper trail; if you bring on teammates, everyone can see what changed, what’s in progress, and what the agents are actually doing without digging through tools all day.

His bigger idea: “non-technical” is dead, and AI spend should pay for itself

From there he zooms out into one of his signature rants: stop saying GitHub is “non-intuitive” or that you’re “non-technical,” because in 2026 you can just ask ChatGPT to teach you the terminology in five minutes. He pairs that with an ROI argument — Claude, ChatGPT, and local hardware like Nvidia’s DGX Spark are expensive, but he says he spends $1,000-$3,000 a month on AI and gets 10x, 20x, even 100x returns by using it to automate research, improve videos, and build businesses.

Hardware demand, Henry, and the big sign-off

Late in the stream he pulls up a chart showing how tiny the population of heavy AI users still is, then makes the case that even one small “red dot” of Claude Code and Codeex users is already causing massive shortages in GPUs, memory, CPUs, and fiber optics. He closes by teasing Henry Intelligent Machines again — first beta starts in about a week, the first tester is already picked — and wraps on a grateful note, still a little stunned that the kid who didn’t think he’d hit 5K on YouTube is now ending a stream at 200K.