In 2024, Sam Altman said a group of tech CEOs had a betting pool:

When would the first one-person billion-dollar company appear?

That idea is beginning to look less ridiculous.

Polsia is one attempt. Its founder, Ben Broca, uses AI agents to plan, code, market, and operate companies.

Polsia had crossed $3 million in annual recurring revenue with Broca as its only employee.

One founder still sits above the system. He sets the direction and carries responsibility when the agents get something wrong.

Tang Jie, founder of Chinese AI company Zhipu (GLM), thinks this arrangement is temporary.

In his internal letter called The Great Wave Has Arrived, Tang described the move from the one-person company, or OPC, toward the no-person company, or NPC.

I am Alex, welcome to ShortCu8 by Innov8.

Lets Dive Deep 🐰

Use four questions to judge any claim about an autonomous AI company:

Each missing ability sends the work back to a human.

AI agents can now write software, research competitors, draft campaigns, answer support tickets, and monitor a product.

This gives one founder the output of a small team.

The founder remains the operating system. They decide what matters, connect information between agents, catch bad decisions, and restart work when an agent gets stuck.

Polsia makes the current boundary visible. The agents may perform the work, but the founder still owns the goal and the judgment above it.

Finishing one coding task is different from running a business for six months.

The system must keep a long objective alive while customers complain, experiments fail, priorities change, and new work arrives.

Tang calls this long-horizon task capability.

A company cannot restart from a blank chat every morning.

Its agents need a usable record of previous decisions, customer history, failed experiments, current priorities, and unfinished work.

Long context and retrieval help. Reliable company memory also needs to preserve why a decision was made and when that decision has become outdated.

Autonomy becomes dangerous when the system cannot recognize a weak result.

The company needs agents that review code, question plans, compare outcomes with targets, and flag contradictions between departments.

Tang describes self-evaluation as one of the abilities now beginning to emerge in frontier models.

Detection is only half the job.

The system must retry safely, change its plan, ask for missing information, or escalate the decision when confidence is low.

Without recovery, a no-person company is simply an error running overnight.

Tang's letter is bigger than a startup prediction. It is Zhipu's view of the road to AGI.

The company's new two-year "Touch High" plan focuses on four areas:

The first two make the no-person company possible. Self-training is meant to help models produce data, improve code, and accelerate their own development. Safety determines how much authority these systems can receive.

The letter goes further into ASI, emotion, and consciousness. Those parts are Tang's forecast. The near-term claim is easier to examine: AI companies are trying to replace individual tasks with a continuous operating system made of agents.

Zhipu is pairing this plan with GLM-5.2. Its released weights carry an MIT licence, and Zhipu reports a one-million-token context window.

That makes the model portable enough for other companies to download, host, modify, and build upon. Reproducing the complete model would still depend on access to its training data and full training pipeline.

This is part of Tang's larger argument: advanced intelligence should remain buildable outside a small group of model providers.

Sam Altman's one-person-company is a reachable dream now. Polsia offers a current attempt, although its revenue figure comes from its investor and should be read with that attribution.

None of this proves that a no-person company can operate reliably today.

It shows where labs and founders are trying to go next.

Now go build something great!

We read every reply - just reply to this email and let us know how we can improve !

#AD1

Speak your PR description, bug reproduction, or Cursor prompt. Wispr Flow auto-tags file names, preserves variable names, and formats everything for immediate paste into GitHub, Jira, or your editor.

No re-typing. No context gaps. No mangled syntax. Works natively inside Cursor, Warp, and every IDE at the system level.

4x faster than typing. 89% of messages sent with zero edits. Used by engineering teams at OpenAI, Vercel, and Clay.

Try Wispr Flow free

#AD2

Skip the sales call. Walk through Gladly's interface yourself — the AI suggestions, the unified customer view, the full conversation thread. 15 minutes, no installation, no commitment.

See it yourself