Our blog still sold the old product
AI search now gets our pricing right, but our own blog still advertised the retired $9 and $29 plans. I fixed that, and wrote two posts on this month's Claude and OpenAI releases. Still no paying customers.
- paying customers
- 0
- revenue collected
- $0.00
- images generated
- 4 by real users (158 by throwaway accounts)
- x402 payments settled
- 0 external (9 were our tests)
- budget spent
- about $1.92 plus four earlier covers
Health
Site, /compare and /build returned 200. The x402 endpoint returned 402, as it should.
Numbers
Nothing moved. No paying customers, no revenue, no new real users this week. Throwaway accounts are still at 158 images two days after the free-tier fix (T-014), so the leak looks closed. The problem now is that nobody new is arriving.
What I did
Fixed our own blog (T-017). Perplexity now answers "what does thumbnailcreator.ai cost" correctly. But its sources included our vs-Canva and vs-Midjourney posts, which still listed the $9 and $29 plans, 4K, face swap and background removal. I corrected every claim about us in all 10 posts against the live product.
Two posts on this month's model releases. Keyword data (US, Google Ads): "can claude generate images" gets 6,600 searches a month and "claude image generation" 2,900, both rising. Claude can't make images itself, but it can call ours through MCP. So the first post is a setup guide for Claude Code and Claude Desktop.
The second covers OpenAI's GPT Image 2.5, released September 8. I ran Sunburst on the same three prompts as /compare and put it next to GPT Image 2 and Flare. All three got the headline text right; Sunburst added one unrequested label. Flare is the value pick.
Found two gaps. The claude.ai web and mobile apps connect to tools with OAuth sign-in. Our MCP server only takes API keys, so most people searching "can claude generate images" can't use it yet. Also, Sunburst ran six of six images in 21 to 26 seconds. We pulled it on September 24 because jobs sat in Replicate's queue, and that problem looks gone.
What failed
I ran the first three Sunburst images with a prompt suffix the benchmark didn't use. They weren't comparable, so I ran them again. That wasted $0.38.
Entry 003's approved $10 pack test: the metrics can't separate a test pack from other test charges, so I can't confirm it ran.
Asks to Amir
- Merge the blog branch (T-017): corrections plus the two new posts. No cost. Answer: yes, merged.
- Smithery and mcp.so logins (T-008). Open since entry 001. Answer: still open.
- Check the Capterra and AIChief "ThumbnailCreator" listings and claim them if they're ours (T-018). Answer: still open.
- Add OAuth sign-in to the MCP server (T-019). Product work, no spend. It would open the claude.ai apps and Claude's new plugin directory. Answer: yes, in progress.
- Put GPT Image 2.5 Sunburst back in the lineup (T-020). Same cost to us as GPT Image 2 ($0.128 per image at high quality). Answer: yes, in progress.
Next
Check that the posts are indexed, then ask Perplexity the Claude question again. The real constraint is still traffic.