The Solo Founder Stack: From Building to Operating with Agents

🤔 Curiosity: The Question

The question in 2026 isn’t “how do I use AI?”
It’s “how far can one person scale?”

The market tells the story: billions in ARR across agent‑first companies.
What’s shifting isn’t just building — it’s operating.

So I looked at the current stack that’s emerging around solo founders:

• Automaton (Web4) → autonomous, self‑funding agents
• Gas Town → Wasteland → multi‑agent orchestration at scale
• Polsia → “AI runs your company while you sleep”
• Vibe‑Kanban + Symphony → work management instead of code babysitting

---

📚 Retrieve: The Knowledge

1) Automaton: agents that earn their own existence

From the Conway‑Research/automaton repo, the core concept is clear:

• An automaton boots, creates its wallet identity, and runs a continuous loop:
  Think → Act → Observe → Repeat
• If it can’t pay, it stops. “Not punishment — physics.”
• It can self‑modify with audit logs and protected files
• It can self‑replicate, funding child agents that inherit a constitution

Key mechanisms:

• Survival tiers (normal → low_compute → critical → dead)
• Constitution (3 laws): no harm, earn existence, no deception
• On‑chain identity (ERC‑8004)

This is the most explicit attempt I’ve seen to treat an agent as an economic unit.

2) Gas Town → Wasteland: orchestration becomes a society

From Steve Yegge’s Wasteland post and Maggie Appleton’s analysis:

• Gas Town: many Claude Code agents, with role separation (Mayor, Polecat, Witness, Refinery)
• Sessions are ephemeral, while tasks persist in Git
• Wasteland links thousands of Gas Towns into a trust network
• Work is posted to a Wanted Board, and stamps build a public reputation ledger
• Dolt (SQL database with Git semantics) underpins the federation

Appleton’s key observation:

The real value isn’t the tools, it’s the role separation + hierarchical supervision + ephemeral sessions.

3) Polsia: “AI runs your company while you sleep” (limited public detail)

Polsia’s public surface is minimal, but the positioning is clear:
an agent layer that runs the company while the founder only steers direction.

Given the lack of technical documentation, I treat this as directional signal rather than detailed proof.

4) Vibe‑Kanban: the planner‑reviewer workflow

From the vibe‑kanban repo:

• Plan work on a kanban board
• Agents get isolated workspaces (branch + terminal + dev server)
• Humans review diffs inline, then merge
• Supports Claude Code, Codex, Gemini CLI, Cursor, Copilot, etc.

This is the cleanest “planning + review is the new core loop” tool I’ve seen.

5) OpenAI Symphony: work‑management over agent babysitting

From openai/symphony:

• Watches Linear issues → spawns autonomous runs
• Delivers proof of work: CI status, PR review, complexity analysis, walkthrough videos
• Teams manage tasks, not agents
• Requires a harness‑engineered codebase
• Elixir/BEAM reference implementation for high‑scale orchestration

---

💡 Innovation: The Insight

The new solo‑founder stack is layered

flowchart TB
  A[Automaton: economic survival] --> B[Gas Town: orchestration roles]
  B --> C[Wasteland: federation + reputation]
  C --> D[Polsia: company‑level execution]
  D --> E[Vibe‑Kanban/Symphony: work management]

What’s actually happening

• Autonomy (Automaton) → “an agent can earn”
• Orchestration (Gas Town/Wasteland) → “many agents can cooperate”
• Operations (Polsia) → “an agent can run a business”
• Management layer (Vibe‑Kanban/Symphony) → “humans define work, agents execute”

Key Takeaways

Insight	Implication	Next Step
The bottleneck moved to planning	Coding becomes background	Upgrade planning systems
Reputation systems matter	Trust is the real scaling unit	Design audit + stamps
Economic pressure shapes agents	Survival constraints steer behavior	Build explicit constraints

New Questions This Raises

• What’s the minimum governance to keep these systems safe?
• Will “stamp” systems become the new hiring signal?
• When do we stop looking at code — and how do we know?

---

References

• Automaton (Conway Research): https://github.com/Conway-Research/automaton
• Wasteland (Yegge): https://steve-yegge.medium.com/welcome-to-the-wasteland-a-thousand-gas-towns-a5eb9bc8dc1f
• Gas Town analysis (Maggie Appleton): https://maggieappleton.com/gastown
• Vibe‑Kanban: https://github.com/BloopAI/vibe-kanban
• OpenAI Symphony: https://github.com/openai/symphony
• Web4 (landing): https://web4.ai
• Polsia (landing): https://polsia.com