Hello you dreamers, pioneers and venturers into the void of the future!
I’m behind a week but that’s because some interesting projects came in in the meantime. We have to work sometimes, right?
Alright lets get to it.
What the heck is going on?
It’s hard not to say anything about Clawdbot and Moltbook. There’s been mass histeria about these two new arrivals, and as usual it’s a matter of balance and criteria.
For those who didn’t hear, Clawdbot (now OpenClaw) is a new too made to perform real tasks for you, which can be convenient, but also a huge privacy and cognitive sovereignty issue.
Giving access to an AI agent to your documents, emails and other private files can be really dangerous, specially if the agent is a closed-source, third party one like Claude or even worse, Grok. You are pretty much resigning your privacy without even knowing who’s in the other end of the line, and what will they do with your data. Europeans should be covered by the AI act and GDPR, but remember these companies usually are not in their jurisdiction.
Clawdbot in only 72 hours went for adoration to a nightmare of IP theft, extortion, crypto scams with $CLAWD and security breaches.
The other big mass histeria was Moltbook, a sort of social media for these agents, where they autonomously were publishing news, gossipping about their humans, and other issues. Supposedly it is an “agent only” social network, but many people demonstrated that it is quite easy to impersonate an agent and just post as a human. So that became really fast a 4chan 2.0, or the rise of agentic shitposting.
Probably the most worrying edge of the blade in this agentic frenzy is the creation of another platform: rentahuman.ai
This website allows agents, not people, to hire humans to do things an agent cannot do in the physical world. And in a world that is flagrantly turning its back to ethics (with “AI UGC” where people that don’t exist talk about products they never tried among the most popular scams) this could represent a real danger for the safety of people, like AI-driven digital mafia, where an AI agent can hire a hitman to menace or harm a person, maybe even without any user approving it. Why can’t we have nice things?
These cases should work as a cautionary tale of the risks of blind early adoption. In a world where Kaby Lame hits a 900 million USD agreement for its likeness (imagine what can be done with 900m), it looks like we’re witnessing the fall of collective reasoning, trading our own agency as species for the sake of speed.
Ok now the nice things
On a more technical and creative tone, it looks like it’s going to be a crazy year for 3D and audio. Many interesting releases around them the last weeks
These are the ones that caught my interest and my attention.
Model of the week: Rodin and semantic editing of 3D models
Probably the most promising and novel tech, this tool allows you to edit 3D models semantically, something that was not possible before.
Versecrafter:
A super interesting video world model that converts videos to coherent space through gaussians and allows you to control the camera on that scene. It even has a Blender add on on it.
VerseCrafter: Dynamic Realistic Video World Model with 4D Geometric Control
This can become super useful for not only filmmaking, but also reconstruction of places like showrooms, sports visualization and other fascinating applications.
ACE-step: a foundational music model
A new musician is in town, and finally came an open source music model with good capabilities like singing and instrumentation, lyric2vocal, singing2accompaniment and repainting.
The most powerful feature as always, is the ability to finetune a model, which can be super interesting for artists to embed their own style and creative decisions into the tool.
If you have an idea around one of these new tools let me know, I’d love to collaborate.
You can find the repo here: https://ace-step.github.io/
Alright this is all for now, see you in one or two weeks!
