A bit more on Merry Man and where Phase 3 is at.
I probably left a few of you wondering yesterday what the hell I meant by a privacy focused AI agent, so here’s a little more context.
Think along the lines of what Venice has done for private AI, but built around the Cerebro inference economy and with our own agent sitting on top of it.
The direction with Merry Man is pretty simple. Your research, code, documents and internal context shouldn’t need to become permanent data sitting on someone else’s servers just because you want to use AI.
I’ve been working on keeping the agent layer separated from the inference providers underneath it, so Merry Man isn’t tied to one model either. As we add more models, the agent can eventually work across that inference layer while Cerebro handles the compute, usage metering and credit consumption underneath.
On the launchpad side, I’ve been spending more time thinking about what happens before anything actually goes live.
I don’t want it to feel like another form where you enter a name, upload an image, deploy something and call it an AI launch.
The build flow is starting to take shape around actually configuring what you’re creating first. Giving it instructions, defining how it behaves, setting its capabilities, testing it privately and seeing how it performs before you decide it’s ready for anyone else.
There are also a few things I’m experimenting with around how projects evolve after launch. I think that part is important. Launching should be the beginning of the software, not the end of the product.
Still playing with some of this, so I’ll save the details until I can actually show it.
And I think we’re at that point now.
Enough text updates. They’re boring, and half of this makes a lot more sense when you can actually see it working.
Over the next few days I’m going to start posting previews of what I’ve been building.
First one shortly. 🧠
