@amitisinvesting This leads to more open source models. Yes you still need memory, but overall capital cost drop dramatically when fine tuning an open weight model. OpenAI and Anthropic can’t compete on price since nvidia and Microns margins are insane. We need capital costs to drop or China wins
@NousResearch Hermes agent and @Railway’s MCP is a sick combo for a designer with no backend service experience! Gave Hermes a simple vague prompt to “duplicate existing services in the same project for hosting new Hermes agents for a new client. It basically one shotted it!
Building a whole investment pipeline using Hermes and cron jobs. I will have several agents retrieve relevant information about my stock positions parse and send to a @supabase table for a knowledge base. Starting to learn about vector embeddings for future retrieval of data.
Started putting the @NousResearch Hermes agent through some design workflows using @paper! Building a new port and along the way testing out what's possible! Maybe an 1hr 30mins in or so ran out of mcp request! Launching subagents to build parts of site wrecks mcp requests haha!
Took a few days with hardly any knowledge of powershell, terminal, etc being a designer, but now I have a dope stack:
@Railway running backend services
@OpenRouter powering models (might change)
@NousResearch Hermes agent harness
@OpenWebUI for the front end
Working on animations for my side project by using @hiunicornstudio
I was not satisfied with just adding static text on top of the gradient. I explored many options and ended up with this diffuse / grain animation that matches the whole theme I am going for.