do you ACTUALLY watch clawdbot content?
thinking companion vs. thinking replacement
do you ACTUALLY watch clawdbot content?
A serious question for you that's come to my mind lately, inspired by the Clawdbot stuff:
If a creator tells you they use Clawdbot for everything - why watch it?
What I mean is, the logical extension of things goes like this:
#1 - You claim to use Clawdbot for everything
#2 - Part of that includes writing all of your social posts & youtube scripts
So it stands to reason then that - what?
I'm just watching a "human skin" regurgitate an AI's script?
Clawdbot went out, researched everything that's "trending", analyzed your competitors, wrote the script, plugged in some retention hacks and made up stories, and now you're just sitting there… reading on its behalf?
And if that's not the case, the alternative would just be that they're lying about what they actually use it for vs. what sounds sexy and cool for a social media post?
Either way, seems like a kinda shitty situation to me
And I think that's something that extends beyond Clawdbot
I've got two friends, both of them are relatively successful entrepreneurs
The first one uses AI as a thinking companion
Ideas originate with him, and the AI is just used as a force multiplier of sorts - a socratic companion that sometimes has skills he doesn't (like writing software), but he's ultimately always the one in the driver's seat.
If the AI outputs a shit idea - he looks at it and says "no, that's shit. Think of it more like this…"
The second one uses AI as a thinking replacement
He relies on AI for the act of reasoning about abstract problems, trusts the outputs with 100% certainty, and let's it do whatever the fuck he wants.
If the AI outputs a shit idea - he looks at it and says "full send bro"
I'll let you guess which one has a growing business and which one keeps getting punted in the nuts!
What is the point behind this rant?
Everyone is so focused on attempting to push the limits of what AI can do, that they don't really slow down to ask the question:
Where is this stuff really going to land in a few years? Where should it land?
Do you actually think AI will entirely replace all aspects of human interaction?
It'll create our TV shows, paint us pictures, breakthrough fusion energy constraints, and make great memes?
It's kinda obvious to most of those grounded in reality that AI is here to stay, but that having humans as the driving force behind the creation of new ideas will always be there.
I think it's important to keep that framing as a filter as you move through the next few months/years.
A quick little analogy
Do you remember back when the internet was actually good?
I think my audience here is generally in the 30-50 yr old range, so you probably remember the glory days.
The birth of memes, browser games that are better than modern AAA games, sketchy chatrooms you shoulda never been lurking in.
Then what happened?
Among other things, a rise in things like programmatic SEO (people creating pages automatically to rank for search terms), automated content production, social networks becoming social media - stuff like that started to take over.
And then it all became very empty-feeling, devoid of real human touch.
Sure, back in the "glory days" that human touch might have translated to "guy that clearly lives in his mom's basement with a few too many crusty socks"...
But at least it was fun.
Anyway… the AI space feels like it's at a similar crossroads.
WHAT ELSE YOU SHOULD CHECK OUT
Things I’ve Learned
Claude code's CLI source code got leaked
While I'm sure Anthropic wishes it was an April Fool's joke, they did in fact leak the Claude Code's source.
This doesn't mean the model weights and all of that stuff behind Opus/Sonnet, but the actual Claude Code CLI tool.
If you wanna kick back and have a 'LOL' about it while you learn a few things, check out Fireship's 7 minute video on it:
Spec-first > vibe-first
I personally don't really use Claude Code's "plan mode" unless what I'm doing is already super, super, super clear.
While they haven't really gone anywhere, I think we're gonna see a big uptick in tools like Speckit, GSD, and others.
I heard a quote somewhere that I will now quote as if I had the study sitting in front of me:
"The biggest cause of failure in AI coding is underspecified plans"
Any time I hear someone complain about AI not building something the right way, I always think: "well, the spec must not have been very clear then".
If Linus Torvalds can debug the linux kernel with AI coding tools, you should be 100% able to make Tinder for Cats!
The best way to manage coding on an existing project
I've had a lot of asks in my paid group about tips for managing building stuff inside of established projects.
While there's a handful of things you can do there, one thing that was pretty clear in the Claude Code leaks is that they comment the living shit out of their codebase, and leave no ambiguity.
They'd rather the model consume tokens (in the form of comments) to understand how something works, rather than have it mess up - and I agree!
One step in that direction is something called "intent layers", you can read this blog post if you're interested in how it works:
That's it for this week,
-Sean
P.S. - if you're interested in building, shipping, & getting your first customer, I've got a paid community you can check out here:
https://www.skool.com/tech-snack-pro
It's currently at 67 members, but price is likely going up once I launch v2 later this month.
V2 is gonna cover:
- My full PRO framework for building apps
- Claude Code masterclass
- Shipping, Deploying, & Monitoring
- Customer acquisition stuff (copywriting, pricing, activation models, landing pages, where to launch at and what to say, etc)