
Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Alex Finn’s SaaS thesis is brutally simple: you need distribution, network effects, or proprietary partnerships/data — he says Creator Buddy’s $400,000+ came from audience-led distribution, and argues any SaaS with one of those three can make you a millionaire, while all three together is “billionaire, potentially trillionaire” territory.
He thinks coding is the whole AI game now — Finn says Anthropic became a monster because Claude Code stayed best-in-class, OpenAI finally “figured it out” with Codex, and every lab’s real leverage is whether its model can code, not Gmail features or side projects.
Grok Build CLI surprised him by finally being competitive, but not elite — after years of Grok failing his recurring coding benchmarks, Grok Build scored 27.2 overall: behind ChatGPT 5.5, Opus 4.7, Opus 4.5, and Gemini 3 Pro, but ahead of older Grok versions and GPT 5.1.
The strongest Grok moment was the 3D city fly-through — Finn called it the best city benchmark result yet, praising detailed windows, road layout, and streetlights, though the camera work was so bad it often missed the buildings entirely.
He’s bullish on ChatGPT image generation almost to a weird degree — Finn calls “ChatGPT image 2.0” horrifically underrated, says it might be “the biggest release of the year,” and shows that his recent thumbnails came from basically one lazy prompt: “make me a thumbnail for this.”
He’s doubling down on raw personality over polished production — Finn says AI content is getting too overproduced, wants his second channel to be mostly him “turning on a camera and talking,” and argues his edge isn’t editing but being more interesting than other AI creators.
Finn opens in full preacher mode: software isn’t dead, SaaS is still a path to “tons of money,” and he wants to show exactly how he did it with Creator Buddy, which he says has made over $400,000. He also immediately frames X/Twitter as the place to build if you want to sell software, while joking his way through quarter-zips, bad DMs, and people already complaining that he hasn’t hit the title topic fast enough.
When he finally gives the quick version, it’s three things: distribution, network effect, and partnership/data. Creator Buddy worked because he had an audience; Instagram is still valuable because everyone’s already on it; Palantir matters because of government relationships and unique data. His line is the sticky one: one of those makes you a millionaire, all three makes you a billionaire.
Before touching Grok, Finn goes on a long riff about why coding ability is the only thing that matters. He says Google ships plenty of AI features, but not the ones people care about — frontier model updates and coding tools — while Anthropic understood that from day one and may become “the most valuable company in the world” because Claude Code stayed great. OpenAI, in his telling, woke up, killed distractions, and pushed hard into Codex, which he calls the best vibe-coding experience on Earth.
In classic Finn fashion, he pauses the main topic to change the stream title for “virality” and make a new thumbnail live. The point he keeps hammering: there’s no magic prompt anymore — he just asks ChatGPT to make a thumbnail and gets something he considers fire. He calls ChatGPT image generation “the Opus 4.5 moment for image gen,” which is one of the more intense compliments he gives any product all stream.
Once he finally installs Grok Build, Finn’s first reaction is aesthetic: “that’s a good-looking CLI,” maybe the best-looking one he’s seen. But even in the honeymoon moment he’s annoyed that xAI is charging effectively $300/month for premium access and takes a shot at the industry habit of shipping everything in “beta” instead of releasing polished products like OpenAI and Anthropic do.
He runs Grok through his recurring tests. The 3D first-person shooter initially melts his Mac Studio M3 Ultra by spawning a ridiculous number of enemies, but after one fix he admits it becomes one of Grok’s best-ever results, around an 8.5. The Elon Musk dance animation is a dud — sparse, awkward, missing the detail other models usually add — and he scores it around a 3.1.
The city fly-through is where Grok finally earns real praise: best windows, best street layout, best sense of an actual city, even if the camera keeps looking the wrong direction. The music visualizer gets a more mixed reaction — cool functionality, bad music — and lands at roughly a 7.2. Total score: 27.2, which puts Grok Build in the ballgame, but still behind ChatGPT 5.5, Opus 4.7, Opus 4.5, and Gemini 3 Pro.
The closing stretch is pure Alex Finn energy: a rant about people demanding he “pick a side” between tools, platforms, or creators, and why that mindset makes the internet miserable. He says AI changes every five minutes, so loyalty to any one tool is stupid; he only cares what’s best right now. He also hints that streams may move from three times a week to once on Fridays so he can focus on shipping Henry and running Vibe Coding Academy bootcamps with topics like SaaS selling, AI investing, and Linear/Slack/GitHub workflows.
Share
Keep Reading
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.

Playbook
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.

Playbook
OpenAI shipped /goal for the Codex CLI. It turns a prompt into a persisted, self-continuing contract.