Vibecode Caveman

Looks good, but it doesn’t sound like Harry Potter… John Williams created the foundational sound for the HP universe….and that is getting chucked out the window. Hans Zimmer (Dark Knight Trilogy) / Bleeding Fingers Music is composing an original score for the series.

If you are someone who does this, please step in front of a bus. Thanks.

Psychotic

Me after scrolling through LinkedIn for 4.5 seconds

@bryan_johnson What if the water is contaminated?

if your hands get dirty you can just wash them

Holy smokes @Replit agent 4 is so good!

@MannyBernabe I’ll wear the heck out of that

Who want's this?

Update

@RobMc5y Honestly, not bad. Publishing now.

@VibecodeCaveman This might be…amazing!

Gave Claude Chrome full access to a mobile app project in Replit 😬 We’ll see.

@kritarthmittal Sure, just not in your AI influencer department.

@VibecodeCaveman bro you got a lot of time, wanna work for me? im hiring

We already drown in human influencers peddling garbage, do we really need AI ones? Barrier to entry’s zero, so everyone’s gonna flood the feed with slop. Going to be similar to OnlyFans…top 1% rake it in, the rest starve. Just because we can doesn’t mean we should. The tools we have now are amazing and we can create literally anything. Something for education, art, useful tools…..and we’re churning out fake monks and thirst traps?! Give me a fucking break. And yeah, boomers lap it up. Either they know it’s fake and don’t care, or they think it’s real and get duped. Once they die off, this bubble pops. None of the younger generations are falling for this shit. Waste of compute, waste of time. Build something that you actually care about. Build something useful. Solve a problem…this ain’t it.

@kritarthmittal Is that the prompt you used for this?

@VibecodeCaveman cry harder

Listening to this now! Also… when did listening to articles become a thing?!

Listening to this now! Also… when did listening to articles become a thing?!

@amasad Why remove the top bun if he’s just gonna put ketchup on the side

Warhol remains the undefeated burger eater

Just because we can doesn’t mean we should. Besides, once all the boomers die off, nobody is falling for this anymore.

Been working on setting up OpenClaw for the past week or so, and holy shit, what a ride. What's working: - Memory actually persists between conversations - Web search is enabled (Brave API) - Can research topics and answer questions - Running on Together AI cloud (switched from local models) - Got Squirt (my agent) responding via Telegram - Dedicated Mac Mini setup (took way longer than expected) What didn't work: - Local LM Studio models were a disaster (too slow, memory issues) - Job board scraping is way harder than I thought (LinkedIn blocks everything) What I'm still figuring out: - Actually getting automated job search running (skills exist but setup is technical) - Web scraping for a social media project - trying to automate content creation from scraped data and generate graphics. Anyone done something similar? - I thought it would be as simple as “Hey Squirt, look up XYZ, pull the following information, and compile it into a nice, neat little graphic…” Apparently not. - Whether this whole thing was worth the time investment… jk You can't just install OpenClaw, send it an X post, and say “do this.” At least not right away. It's more technical than some folks make it out to be. But if I can figure it out, anyone can. I wrote about wasted time having exponential cost in the AI age. Turns out the flip side is true too: hours spent learning this stuff compounds fast. Even troubleshooting counts as progress.

@DnuLkjkjh Appreciate the feedback! Going to try building that custom skill then hopefully I’ll be on a similar path. Certainly no shortage of ideas, just have to work out the how.

the jump from local models to a cloud provider is the right call. I run OpenClaw on a Mac Mini too — cloud API for the heavy thinking, local models only for embeddings where latency matters more than capability. for scraping + graphics: build a custom skill with a shell script that does the scraping (curl + jq), then passes structured data to the agent. trying to get the agent to drive browser scraping directly is where most people get stuck. the compound learning is real. first week was all debugging auth and config. now I have cron jobs running twitter, email, and content pipelines autonomously.