Based Sid
136 posts
@0xBasedSid
Managing a swarm of AI agents that are building boring internet companies. Currently deploying @virlfm
We're ready to release the test environment for our agentic infrastructure. In the upcoming hours the status page will be also released with a brief tutorial on how to use it.
Pilot launch of the agentic infra is set for tomorrow. Initially, only one agent (our own) will be deployed to verify the correctness of VIRL FM development. If the launch goes well, the feature will be rolled out to a limited number of users first.
If someone says they can launch a token on Base or Solana easier, they're lying. Introducing virl.fm. Agent-native launch infrastructure for tokenizable content, businesses, moments on Base and Solana Live on pumpfun: 2SpvcXGq6YYPK5EWPSVawDsLWBAXGuE6uQuNLb3dpump
We are building this as infrastructure, not a single closed app. It is multi-tenant from the start: API keys, usage tiers, metering, and a trading-fee rail injected directly into Jupiter routing on Solana. A public SDK is in development so any developer can run agents on the same rails we use internally. virl.fm runs as the first consumer of that platform, not a special case. If we can't build on it, neither can anyone else. Where this honestly stands today. The foundations, the single-agent execution path and the multi-tenant core are already written and type-checked across the codebase. What is left before a closed beta: - live testing against funded wallets - full Base and Solana strategy parity - gasless agent launches - a marketplace of strategies and composable skills We'll keep posting the build as it ships, not after.
What drives an agent is not the chat box. Under each one is a position state machine we are porting from virl's live launch-and-sell bot: take-profit, stop-loss, time-cap and graduated exits, re-evaluated on every tick against real market data (RSI, EMA, momentum). You choose a strategy: monitor-exit, momentum, sniper, defi-yield or macro-guard. The model can propose a move, but the state machine and the policy engine decide what is actually allowed to execute.
The first thing we're solving is custody, because most "AI agents" quietly hold your keys. Ours don't. Each agent gets its own wallet, but signing happens inside a secure enclave (Turnkey), behind a policy engine: an address whitelist plus hard spend-caps enforced at the signing layer, not in our backend code. The agent can trade within those limits. It cannot send funds to an address you never approved, and it cannot spend past the cap. Non-custodial because of how it is built, not because we say so.
We're building an agentic trading layer for $VIRLFM. You deploy your own autonomous trading agent, fund its wallet, and it runs a strategy on Base and Solana on its own. The same friction we took out of launching a token, taken out of running a position. This is in active development right now. It is not live yet, and we won't pretend it is.
While I'm testing how's everything working, I already started on a fully agentic environment for token launches. Beta may go live by the end of this week. And this feature will require to hold a portion of $VIRLFM
Thrilled to announce the @bankrbot integration into VIRL FM. From now on, you can deploy tokens via Doppler, Clanker, Bankr and Pump fun using our app. P.S. This is still a beta version, so bugs may occur.
