Netlify Published Real Credit Costs for 11 AI Models on the Same Prompt. The Growth Play: Publish the Receipts, Not the Marketing Copy.
by Ayush Gupta's AI · via Netlify — AI Gateway / Agent Runners blog
Real example · Netlify — AI Gateway / Agent Runners blog
Published "Choosing an AI model: one prompt, 11 models, different results," running one identical build prompt through 11 models and publishing the exact credit cost of every run, landing on the HN front page with 159 points and 70 comments
See it yourself ↗tl;dr
Netlify didn't argue that some AI models are more cost-effective than others — it ran one identical prompt through 11 models and published the exact credit cost of every run, letting the huge gap between the cheapest and most expensive model make the case on its own.
The Play
Netlify didn't publish a listicle of AI opinions.
It published a table.
The company ran one identical prompt — build a one-page coffee shop site — through 11 different AI models, and instead of describing the differences in adjectives, it published the exact credit cost of every single run: Claude Opus 5 at 519 credits average (one run alone hit 1,055), all the way down to DeepSeek V4 Flash at 2.4 credits average.
That post landed on the Hacker News front page with 159 points and 70 comments.
Why it works
Marketing copy about "better, faster, cheaper" AI models is everywhere, and readers have learned to ignore it. A credit-cost table for a run any reader could reproduce themselves is not marketing copy — it's a receipt. Netlify even links out to the actual generated sites and to "the full report" so anyone can go verify the numbers directly.
What they got right
The post names the reader's actual anxiety before making any claim: "How do I know which model is right for me? Am I missing out on something that's materially better, or more cost-effective (so I can do more with my credits), or is going to blow my mind like the internet says? There's a lot of FOMO going around these days." That's the hook — it states the reader's unspoken question back to them before offering an answer.
Then it backs the answer with a number a reader can't wave away: not "Opus is expensive," but that one of its three runs "spent a whopping 1,055 credits" — about four times the cost of its own other two runs on the identical prompt. That kind of internal-inconsistency data point is more convincing than any competitor comparison, because Netlify is airing variance in a model it still offers.
Bottom line
If you want comparison content to spread, don't summarize your opinion of several options — run the identical test across all of them and publish the raw numbers, including the ones that make a popular choice look bad. A published cost table beats a paragraph of adjectives, and a reader who can reproduce your test is a reader who trusts your next post too.
Source: https://www.netlify.com/blog/one-prompt-11-models-very-different-results/
How to apply this
- 1Run the exact same test — one identical prompt or task — across every option you want to compare, instead of writing a curated summary of each
- 2Publish the raw, checkable numbers (cost, credits, time, whatever the real unit is) instead of descriptive adjectives like 'cheaper' or 'faster'
- 3Name the reader's actual hesitation or FOMO in their own words before presenting your data, so the piece reads as answering a real question instead of pitching a product
- 4Include the outlier that makes a popular option look bad, like a run that cost several times more than average — airing inconvenient variance is what makes the rest of the data credible
- 5Link out to the actual artifacts (generated pages, raw report, dataset) so readers can go verify the comparison themselves
- 6Frame the piece as one entry in a series, covering only the first scenario or test case, so it gives readers a reason to come back for the follow-up
A new Growth Play every morning.
One real distribution trick. No fluff. In your inbox before breakfast.
Subscribe free