Praneeth got his first users primarily through Twitter (X), but not via direct product promotion. Instead, he replied to existing conversations related to AI interactions and the "new age of AI," hinting at problems his product solved rather than pitching it outright. He was doing "build in public" posting but acknowledged he didn't have a large personal brand, so the organic reach was limited. He also made sure his landing page H1 and copy explicitly called out "power AI users" as the target, which helped convert the visitors he did drive there. The critical unlock for crossing 100 paying users was switching from a free-trial subscription web app to a one-time purchase desktop (Electron) app with a bring-your-own-keys (BYOK) model. He launched at $29 as an introductory lifetime deal price. This low price point lowered the barrier enough that users who were already subscription-fatigued from paying for multiple AI tools (ChatGPT, Claude, Perplexity) were willing to make a one-time commitment. He explicitly priced at $29 until he hit 100 users, then bumped to $49.
Canvas Mode (formerly Rabbit Holes AI)
Canvas-based desktop app for power AI users to have branching, non-linear AI conversations
7 moves, in order
- Initial Launch (Web App, Subscription)Build in public — Twitter
Launched a web app with a free trial (limited free questions before paywall) and posted about it on Twitter as part of a build-in-public effort, replying to AI-related conversations.
Got some interest but zero meaningful revenue; no product-market fit. - Pivot DecisionProduct repositioning
Identified that his ICP (power AI users) was already paying for multiple AI subscriptions and had subscription fatigue. Decided to switch to a one-time purchase model and eliminate all server/API costs by building a BYOK (bring-your-own-keys) Electron desktop app.
Eliminated ongoing server and AI API costs, reducing monthly overhead to ~$50 (Vercel hosting) + $200 (Cursor subscription). - Re launch as Desktop LTD (0–100 users)Lifetime deal launch
Launched the Electron desktop app as a lifetime deal at $29. Priced intentionally low to validate demand and attract early adopters. Landing page H1 explicitly called out 'power AI users' to speak directly to the ICP.
Crossed 100 paying users; buyers were more serious, gave better feedback, and were more engaged than free-trial users.Users 100 users - Post 100 UsersLifetime deal price increase
After crossing 100 users, raised the lifetime deal price from $29 to $49. Used the milestone as a lever to create urgency for fence-sitters.
Continued user growth; used pricing ladder as a growth signal and revenue accelerant. - Growth PhaseTwitter community engagement
Replied to Twitter conversations about AI interactions and next-generation AI tools — not direct product pitches, but adjacent commentary that hinted at the problems Canvas Mode solved. Also ensured landing page copy was tightly targeted to power AI users.
Drove traffic; site reached 38,000 unique visitors total. - 6 Month MarkLifetime deal overall
Continued selling LTD (60% of 1,200 users are lifetime buyers) alongside an annual one-time purchase option (40% of users, requiring renewal after 1 year for updates). BYOK model kept margins near 100%.
$80,000 gross revenue, ~$75,000 profit in 6 months with 1,200 paying users and 38,000 unique site visitors.Users 1.2k users - Next Phase (Planned)Subscription model transition
Planning to phase out or raise price on the LTD and transition toward a recurring subscription model now that they have a validated user base of 1,200 paying customers to support the shift.
Not yet executed; stated intent only.
Prior UX design background helped him identify and articulate the problem clearly. He self-funded from savings after quitting a $250K/year software engineering job, giving him runway to iterate without investor pressure. He also used Cursor + Vercel AI SDK heavily to accelerate solo development.
lifetime_deal_pricing_with_byok
Free-trial subscription (freemium) web app model got interest but "never really turned into any real dollars." Build-in-public Twitter posting without a large existing personal brand also failed to generate meaningful revenue. Server-side hosted model was also abandoned because ongoing AI API costs and server costs made the economics unworkable for a bootstrapper.
My app failed, then I changed one thing, and made $80K
Jenny AI
AI-powered writing and research assistant tools for students and researchers
Chatbase
AI chatbot builder — train a ChatGPT on your own data/documents
StealthGPT
AI tool that makes AI-generated text undetectable by AI detectors