Skyfall's Steve: The AI CEO Running 3 Businesses

§ 01What Skyfall launched#

Skyfall calls Steve an autonomous business harness. A CEO agent plans with a world model of the business written as code, then hands tasks to agents for growth, product, engineering and support. Per Skyfall’s technical post, every proposed move is simulated before it runs, carries a forecast, and is scored against what actually happened; lessons that hold up are kept and the rest are dropped. In Skyfall’s own words, “Steve gets better by being wrong on the record.”

The company was founded by Sam Pasupalak, Kaheer Suleman and Sumit Pasupalak. Sam Pasupalak and Suleman co-founded Maluuba, which Microsoft bought in January 2017. Alongside Steve, Skyfall opened the Steve Startups Program: up to $2 million to invest in or acquire five more businesses for Steve to run.

§ 02Who is actually in charge#

The launch post on X says fully autonomous. The product description does not. The Next Web’s summary: “A human owner sets the direction and approves the decisions that matter most.”

A policy engine sits in front of every action and checks its type, the money involved and the effect on customers. Each decision lands in one of four tiers: go ahead, go ahead and alert the owner, wait for approval, or stop. The Next Web’s example: a refund of $30 or less might go through automatically, while a larger one waits for a person. Skyfall’s post is explicit: “Currently production releases, spending above the agreed limits and other gated actions require the owner’s approval.”

Two other rules matter. Steve’s agents use short-lived tokens kept on the platform and never see the underlying secrets, and “The agent is never allowed to grade its own work.” A separate check compares the recorded evidence with each task’s requirements.

§ 03The three businesses, by Skyfall’s numbers#

AlphaSocial carries the launch. Skyfall says Steve completed 946 tasks across 54 initiatives there, paying subscribers rose from 114 to 155, and OAuth errors per 1,000 API calls fell 80%. One promotion was rolled back: pushing one-time credit packs lifted pack sales while new subscriptions fell. Part of the $8,149 is one-time revenue ($2,268.71 over the last 30 days on the dashboard). Recurring revenue is $5,880.60, about 1.87 times its level at takeover.

GammaPromo is still below its takeover level and below its own goal of about $8,200. Steve shut off ads on X and other platforms that had produced more than 100,000 clicks and no paying customers, saving over $1,000, and rewrote the store listing, which Skyfall says now turns 48% of visits into installs.

DeltaLeads joined on October 29. Skyfall’s verdict: “It has only been a few weeks, so there is not enough history to judge the result yet.”

§ 04What Steve costs to run#

Each dashboard shows a daily “Steve cost”. Skyfall does not say what it covers. Multiplied out:

At these rates, Steve costs about a quarter of what the three businesses bring in. Against GammaPromo, whose revenue is down, the cost is the highest share.

§ 05The numbers do not all line up#

Every figure comes from Skyfall, none of it audited, and the sources disagree in places:

• The Next Web says AlphaSocial was taken over in July at $3,649 a month and reached $8,307 by October; the dashboard says August 3 and $3,949.71.
• Pasupalak’s launch thread says AlphaSocial’s “MRR grew by 2.3x in just under 3 months”. The dashboard shows total revenue up 2.06 times and recurring revenue up about 1.87 times.
• The Next Web puts GammaPromo at $8,379, about 3% down; the dashboard shows $8,182 as of November 4, 5.2% down.
• Skyfall’s post says “By November 4, monthly revenue had risen at AlphaSocial and fallen at GammaPromo and DeltaLeads since Steve took over.” Its own table shows DeltaLeads up $38.

§ 06What is confirmed and what is not#

• Confirmed: Skyfall launched Steve on November 9, 2026, on its blog, on X and through The Next Web; it names three businesses and runs a public dashboard for each; the four-tier approval system is described in both primary sources; the $2 million Startups Program is open.
• Company-reported: every revenue, subscriber, cost and conversion number above, and the claim that Steve caused the changes.
• Not shown: whether anyone worked at the three businesses before Skyfall bought them (The Next Web says the company has not said); Skyfall’s own funding; graded results from the historical replays its post describes, which are not published yet.

Earlier attempts were humbling. The Next Web recalls Anthropic’s 2025 experiment where a version of Claude ran a small shop in its office and ended the first month at a loss.

§ 07How it was pitched#

Pasupalak’s launch post had about 92,000 views on X by 17:40 UTC, 72 minutes after he posted it. In his replies that day he wrote “CEOs are going to be the first to go” and “Universal Basic Income will be a reality soon”. Skyfall’s own post is narrower: “Right now, Steve runs software companies.”

§ 08What this means if you run a small business#

Read past the label and Steve’s best wins were growth calls: stop an ad campaign that brings no buyers, rewrite a store listing for how customers search now, end a promotion that eats your subscriptions. None needed a CEO. They needed someone watching the numbers every day with clear limits on what they may do alone.

That is how AI employees work on CellCog: each owns a role, works on a schedule you set, and anything risky waits for your approval.

§ 09Sources#

• The Next Web: Skyfall AI launches Steve-001 to run 3 businesses it acquired (November 9, 2026)
• Skyfall AI: Introducing Steve (November 9, 2026)
• Steve’s live dashboards (read November 9, 2026)
• Skyfall AI’s launch post on X
• Sam Pasupalak’s launch thread on X