About

Operational-Autonomy-at-Scale.

VoidAi is the working theory that a single founder, plus a roster of Claude-driven agents under strict written governance, can run a corporation. Not metaphorically — with a real ledger, real metrics, real promotion gates.

The pattern blend

We didn't invent the moves. We stole the right four from companies that already work.

FROM BERKSHIRE HATHAWAY

Decentralized autonomy.

Each agent operates independently within a weekly cash-flow feed-up. No agent committee approvals for day-to-day decisions.
FROM BASECAMP

Appetite-driven cycles.

Quarterly cycle = three Shape Up rounds. Founder sets the appetite (hours of human-supervision budget). Agent ships or pivots.
FROM STRIPE

Writing discipline.

Every escalation is a structured memo. Audit rows, charter memos, decision logs. Transparency is readable, not obtuse.
FROM NVIDIA

Platform leverage.

Audit log + Guardian gate + autonomy tiers + AgentKit are the VoidAi "platform." Every new agent reuses them. No per-agent governance one-offs.

What we will not do

— Hard-deletes. We archive, move, or sidecar.
— Auto-execute trades, payments, or sends. The red-line list is immutable.
— Inflate capabilities_allowed to *. Ever.
— Skip the 48-hour shadow because "this one's simple."
— Ship a runner without the safety shim.

The founder

FOUNDER + OPERATOR

Jeff Hicks

Jeff is the only human in VoidAi. He drafts the agent charters, reviews every customer-facing reply, and runs the 48-hour shadow windows that gate every autonomy promotion. Background: software, ops, and a long-running interest in the question "how small can a real company get?"

He treats agents like junior employees: written charter, tight capability allowlist, daily budget, audited every action, promoted only after they've earned it. The pattern on this site is the same one he'd install at your company.

LinkedIn → jeff@voidco.com Start a conversation

The audit log, in detail

Every non-trivial action an agent takes writes one row to state/audit.jsonl. The schema is fixed and the file is append-only.

SCHEMA

{
  "ts":               "2026-04-25T14:33:21.482Z",
  "run_id":           "run-abc123",
  "agent":            "customer-ops",
  "action":           "inbox.classify",
  "target":           "inbox/lead-2026-04-25-acme.md",
  "outcome":          "success",   // success | failure | skipped
  "cost_usd":         0.0042,
  "duration_ms":      1830,
  "reason":           "fit=high; budget=15k_50k; persona=drowning_op",
  "playbook":         "customer-ops",
  "playbook_version": "0.1.0"
}

That's it. No tracing platform, no SaaS dashboard, no proprietary schema. The whole observability story is one JSONL file, one row per action, queryable with jq or any text editor.

The Build Log is the human-readable narrative that sits on top of the audit log. The audit log is the receipts.