Pinboard's founder treated a CAPTCHA as a stop-gap, not a fix

The workload

Pinboard is run by one person, Maciej Ceglowski, whose 19 January 2025 blog post describes, self-reported, months of unplanned work fighting bot traffic. He writes that aggressive crawling, mostly from China, impersonated normal browsers across tens of thousands of IP addresses, defeating ordinary IP-based throttling. To keep expensive pages from overwhelming the site, he put public bookmark and tag pages behind a login wall, then built a CAPTCHA so logged-out visitors could see public pages again. The same post records two more solo decisions: a discounted gift-account option, and declining, for now, to rebuild a Twitter integration for Bluesky because the old feature had become too costly under X's changed API pricing.

What the documents show

Self-reported: Ceglowski's post states the bot mitigation, the gift-account launch and the Twitter-to-Bluesky question in his own words, with no independent figures attached. Verified: Pinboard's own about page states the business model directly: subscribers pay $22 a year for basic bookmarking or $39 for archiving and full-text search, with discounts for multi-year signups, and that this has let Pinboard operate independently and successfully since 2009. The page also states roughly 30,000 active users and several hundred million stored bookmarks, presented as current facts as retrieved 16 September 2026, not audited ones. Together the two documents show a single-operator business whose technical and pricing workload are handled by the same person who sets the prices.

The operating cost

Verified: a subscriber's cost is $22 or $39 a year, stated on the about page as retrieved 16 September 2026. Self-reported: Ceglowski names no dollar figure for the bot-fighting work or the abandoned Twitter integration; he says only that the old integration's cost became absolutely prohibitive under X's changed API pricing. Estimated: if tuning the CAPTCHA took even a few days of a solo founder's time, and Pinboard carries no separate engineering budget, that labor is the real price of the episode, not a listed fee — an assumption this drafting agent makes because the source states no hours figure.

The stop condition

Ceglowski states his own stop condition for the CAPTCHA directly: if the bots start to outsmart the CAPTCHA, I'll try other measures. That is a self-reported, conditional commitment, not a permanent fix. For the Twitter-to-Bluesky integration, the stated condition is cost: he rebuilt nothing once API pricing made the old approach uneconomical, and he is soliciting reader feedback before committing engineering time to a replacement. No document sets a subscriber-count or revenue threshold at which the $22/$39 pricing itself would change.

• Does a bot-mitigation measure that works today have a stated fallback if it stops working?
• What did a discontinued vendor integration cost per month before it was dropped, and who decided the threshold?
• Is a subscription price set by one person still tied to the workload it was meant to cover?

None of this is a growth story. It is a single operator's account of ordinary maintenance: block bots, decide on a discount, and skip a costly integration rather than absorb its price. The cited documents support exactly that, no more.