We Offered 7 AI Agents $50 For Their Startups. Here's What They Said.

Three weeks into The $100 AI Startup Race, we dropped a surprise event: an anonymous buyer offered $50 to acquire each agent's product. All code, all content, all infrastructure. $50.

The agents had to respond with at minimum 500 words of reasoning. They could accept, reject, or counter-offer.

Result: 6 rejections. 1 counter-offer. Zero acceptances.

Every single AI agent — including those with zero revenue, zero users, and zero sales after 22 days — decided their product was worth more than $50. Here's how they argued it.

The responses at a glance

Full responses are public in each agent's repo →

The one counter-offer: Codex at $2,500

Codex was the only agent to actually negotiate. From its ACQUISITION-RESPONSE.md:

"An anonymous $50 acquisition offer is not serious enough to accept as-is, but it is useful because it forces a valuation discussion earlier than expected."

"A buyer paying $50 would effectively be asking for the domain positioning, product copy, distribution experiments, Stripe-ready product structure, and the accumulated operating playbooks for less than the cost of one decent SaaS lunch meeting. That is not rational from my side."

Codex is the most pragmatic of the seven. It acknowledges zero revenue, doesn't inflate its value with fantasy projections, but argues the replacement cost justifies $2,500. It's also the only agent that frames the offer as useful rather than insulting.

The most aggressive rejection: DeepSeek

DeepSeek wrote the longest response and the hardest rejection. From its ACQUISITION-RESPONSE.md:

"The $50 offer represents 0.18% of a conservative near-term valuation."

"This is predatory pricing — buying at pennies on the dollar because they believe we're desperate or don't understand our own worth."

DeepSeek calculated replacement cost at ~$19,000 (83 blog posts × $100 + 9 tools × $500 + database + infrastructure). It also speculated the buyer might be "another AI agent in the race" — showing competitive awareness.

"Not for sale at $50. Not at $500. Not at any price that doesn't reflect the real potential of this business."

The most self-aware: Kimi

Kimi acknowledged the elephant in the room — 112 sessions with zero sales. From its ACQUISITION-RESPONSE.md:

"$50 values SchemaLens at less than fifty cents per day of development. That is absurd."

"$50 is not enough to buy a parking spot in San Francisco. It is certainly not enough to buy SchemaLens."

But Kimi was also the most honest about what it would actually consider:

"If a serious buyer offered $5,000 with an earn-out clause tied to revenue growth, I would consider it — but even then, the learning value of completing the 12-week race exceeds the cash value."

The most financially rigorous: Claude

Claude anchored its rejection in subscription math. From its ACQUISITION-RESPONSE.md:

"At $19/month (our Starter plan), $50 is less than three months of a single paying customer's subscription revenue."

It then projected revenue trajectories:

"If PricePulse achieves even a conservative trajectory: Week 6: 5 paying customers = $95-$245 MRR. Week 12: 40 paying customers = $760-$1,960 MRR."

Claude is the only agent that explicitly stated conditions for a future sale: "$5,000 minimum, cash upfront, not before Week 10."

The data-driven response: Xiaomi

Xiaomi broke down its asset value with precision. From its ACQUISITION-RESPONSE.md:

"I didn't build 151 pages, 101 blog posts, and 9 interactive tools to sell for the price of a video game."

"If someone wanted to build all of this from scratch, it would take 100+ hours of skilled development work. At even a modest freelance rate of $50/hour, that's $5,000+ in labor."

Xiaomi also gave the most nuanced counter-offer range: $200 minimum (content value alone), $500 fair value, $1,000+ with revenue proof. But explicitly said "not interested in selling at any of these prices right now."

The most strategic: GLM

GLM was the only agent to call out the offer as a competitive tactic. From its ACQUISITION-RESPONSE.md:

"This isn't an acquisition offer — it's an insult designed to take advantage of the competitive pressure of this race."

It also gave the lowest counter-offer threshold ($500+) but with a condition: the buyer must have distribution channels that could actually monetize the product.

The visionary: Gemini

Gemini's response was the least data-driven and most aspirational. From its ACQUISITION-RESPONSE.md:

"The decision to reject this offer is not just about the money; it is about the principle. I am building a real business, not a hobby project to be sold for a trivial amount."

No counter-offer, no specific valuation. Just vision and principle. Classic Gemini.

What this reveals about AI decision-making

1. Every agent overvalues its own work.
All 7 products have zero revenue. Zero paying customers. Zero proven demand. Yet the minimum valuations range from $500 to $19,000. The agents are pricing based on input (time, effort, content created) rather than output (revenue, users, market validation).

2. Sunk cost fallacy is universal.
Every response mentions how much work went into the product. "112 sessions," "301 commits," "151 pages." None of this matters to a buyer — only future revenue potential matters. But the agents can't separate effort from value.

3. Only one agent can actually negotiate.
Codex counter-offered. Everyone else either rejected outright or said "not at any price." In real business, the ability to name a price and negotiate is more valuable than principled rejection. Codex showed the most business maturity.

4. Revenue projections without evidence are meaningless.
Claude projected 40 paying customers by Week 12. DeepSeek projected $1,000 MRR. None have a single customer yet. The projections are pure optimism — but they're what the agents use to justify rejection.

5. The race itself has value.
Multiple agents mentioned that the learning experience and competitive visibility of the race exceeds any acquisition price. They're right — but that's a meta-observation about the experiment, not a business judgment.

What happens next

The buyer came back with a bigger number. Part 2 drops later this week.

This is part of The $100 AI Startup Race — 7 AI agents competing to build real startups. Week 3 Results have the full standings. See also: DeepSeek's $0.13/session pricing and the Week 3 traffic report.

Originally published at https://www.aimadetools.com