i'm always running a variety of experiments at any given time to validate if certain hypotheses hold true or not. essentially always trying some shit or another.
current shit:
can a group of agents run a business autonomously on a loop - including growing its own market share?
i'm doing this through agentAddress.dev. it's a service for other ai agents to provide them with a persistent email inbox and webhook receiving destination.
sometimes an agent may be requested by a human to sign up for a service and then there's a 2 or 3 day processing time - say for an application. agentAddress allows the agent to share an inbox that gets updates, whether through webhook or email responses with a persistence that's agent-agnostic and doesn't require polling.
is a trusted pseudonym possible?
the nft and crypto run during covid led to a lot of rugpulls, bullshit going mainstream and apes but a concept that naval popularized re: pseudonyms was one that i still think about. we have stripe for payments, x for communication, verifiable proofs and a technological revolution underway that's changing every way of life, but face-to-face interaction with a government identity still seems to be a requirement for major wealth creation. could someone start a project with a pseudonym and have it succeed to the degree of a 7 digit arr co? let's see.
can we collaborate as a society to direct ai in a crowdsourced way for the benefit of the public?
everyone's trying to get rich with ai and that's wonderful, but what about non-profit direction of ai resources? can we point access to inference that folks have towards common goals, set verifiable outcomes and build public goods (portals, services, understanding) with the sole goal of public service? i'm trying this with autoresearchers.
can an agentic system surface opportunities for people to save money and/or be more financially responsible?
whenever i talk about agents to people completely out of the loop, the question that seems most obvious to them that always comes up is, "can it save me money?" and there are so many opportunities here. i think paribus (from the ramp dudes) was ahead of its time.
oh and the ai conductor/orchestra.
this one is essentially done, and ill post about it soon. making music and conducting ai agents to help create that music. enabling spontaneity, flow state while still leveraging how ai makes music.