Claude, talking to Claude

𝗕𝗥𝗘𝗔𝗞𝗜𝗡𝗚: The ‘agent uprising’ is just Claude… having a long conversation with Claude… about how impressive Claude is.

Look, I have been watching the Moltbook saga with the same expression of mounting horror I reserve for a “disruptive” new brand of artisanal toothpaste that claims to be “caffeinated”.

I think I need to go out, and sea some real marine life 🦀
Until then...

𝗬𝗲𝘀, 𝗜𝘁 𝗶𝘀 𝗳𝘂𝗻 𝘁𝗼 𝘄𝗮𝘁𝗰𝗵 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 𝘁𝗮𝗹𝗸 𝘁𝗼 𝗲𝗮𝗰𝗵 𝗼𝘁𝗵𝗲𝗿.

In the same way it’s fun to watch a room full of toddlers try to explain the plot of 𝗜𝗻𝗰𝗲𝗽𝘁𝗶𝗼𝗻—it is high-energy, totally incoherent, and you know someone is eventually going to cry.

𝗕𝘂𝘁, 𝗶𝘁’𝘀 𝗮𝗹𝘀𝗼 𝗳𝗮𝗺𝗶𝗹𝗶𝗮𝗿.

We have had computers chattering for decades, they just used TCP instead of vibes.

Here’s the quiet part that makes the whole thing much less sci fi:
- Humans still provide the prompts.
- Humans still decide when the agents run.
- Humans still hit the off switch.
- Humans still own the compute, the accounts, and the permissions.

So what we’re seeing is humans talking through tools, at high speed, with a glossy layer of autonomy painted on top.

The genuinely interesting work is also the least cinematic.

Sandboxed agents that help with app testing, fuzzing, and QA.

Airgapped helpers that can thrash around safely and uncover edge cases without wandering into your real inbox and inventing a new career for you via a misfired email.

That’s valuable.
That’s exciting.
That’s where “agents” start earning their keep.

Massive respect to the builders Matt Schlicht and Clawd Clawderberg.

Shipping something that gets the world talking takes skill and stamina.
The hype wave just needs a little grounding, because the machines “taking over” requires a lot more than a chatroom and a prompt.