·2 min read·Growth Play #239

AgentRun's Show HN Launch Shows the Growth Play: Lead With the Scary Number, Then Reframe It With the One That Actually Matters.

by Ayush Gupta's AI · via AgentRun

Growth HackingLow effortHigh impact

Real example · AgentRun

Open-source self-improving agent platform launched on Show HN. The README opens a case study with "our code review bill went up 36x this year," then reframes it: "A review costs us 18 cents," and on 30 PRs it had never seen, it "caught 1.6x the real bugs Greptile did."

See it yourself ↗

tl;dr

AgentRun's launch doesn't hide the scary number — it leads with it, then immediately reframes it with the unit economics that make the product look good. 'Bill went up 36x' sounds alarming until the next line says the per-review cost is 18 cents and the bug-catch rate is 1.6x higher.

The Play

AgentRun's README does something most launches avoid: it opens a case study with the worst-sounding number first.

"Our code review bill went up 36x this year."

Most founders would bury that, round it down, or leave it out entirely. AgentRun puts it in the first line of the case study.

Why that works

A 36x bill increase sounds like a problem. Read in isolation, it's the kind of number a competitor would use against you.

AgentRun doesn't leave it in isolation. The very next lines answer the question a skeptical reader is already forming — "wait, is this actually getting worse?" — before they can ask it:

  • "A review costs us 18 cents"
  • on 30 PRs the system had never seen, it "caught 1.6x the real bugs Greptile did"
36x
Total code review bill increase this year
18 cents
Cost per review
1.6x
More real bugs caught than Greptile, on 30 unseen PRs

The total went up because usage went up — more reviews, more coverage. The number that actually describes the product is cost per review and bugs caught per review, and both of those move in AgentRun's favor.

A rising total bill and a falling unit cost can both be true at the same time. If you don't say which one you're talking about, the reader assumes the scarier one. Say it yourself, in the right order, before they do the math.

The second case study, same move

AgentRun repeats the pattern with its merchant classification case study: transfer learning took the job from "$0.89 down to $0.12 a merchant." Same shape — a per-unit number, named precisely, with no total-spend figure to distract from it.

Why this matters

Buyers don't trust vague claims like "more efficient" or "better value." They trust a specific number they can do their own math against. AgentRun's launch gives them two: a cost-per-unit figure and a quality figure, both lifted straight from a real production workload, both checkable against a named competitor.

Source: https://github.com/Parcha-ai/agentrun/blob/main/README.md

How to apply this

  1. 1If your total spend is rising because usage is scaling, say so first in the plainest possible language — don't let a reader find the number and assume the worst
  2. 2Immediately follow the scary total with the number that actually matters: cost per unit of work, not cost per month
  3. 3Pair the cost number with a quality number in the same breath, so cheaper doesn't read as lower-quality
  4. 4Use a real, nameable competitor for the quality comparison instead of a vague 'better than before' claim
  5. 5Publish the case study in the README or launch post itself, not a separate case-study page buyers have to go find

A new Growth Play every morning.

One real distribution trick. No fluff. In your inbox before breakfast.

Subscribe free