A critical assessment of Wabi’s $55M raise at a $1.25B valuation led by IVP — a Replika-founder company priced as a unicorn in ~11 months, built on a third-party model it doesn’t own, with not a single engagement number disclosed.
Not revenue — Wabi is free with no monetization model, so a revenue multiple cannot be computed. The price is set by founder reputation (Replika, 35M lifetime users) and the 2026 personal-agent capital wave, not unit economics. The PR release cites “strong early demand” with zero demand metrics attached.
Not Wabi’s. Kuyda describes a mini-app as “a UI on top of a prompt”; the user selects the foundation model (ChatGPT or Gemini). Wabi is an orchestration and distribution layer. Its suppliers — OpenAI and Google — are shipping the same “describe-an-app, get-an-app” capability natively.
Unknown. Mini-apps for “date night” or trivia are inherently disposable; the category’s unsolved problem is retention. No cohort of active Wabi groups has been published. Importing Replika’s 35M as reassurance is the central sleight-of-hand — that is a different product’s reach number, not evidence of Wabi engagement.
Key Finding: The round facts are rock-solid — $55M, $1.25B, IVP lead, marquee syndicate — but everything that would justify the price is absent. The honest frame is not “AI agent hits $1.25B”; it is “Replika’s founder gets a unicorn price before publishing a single engagement number,” on a product built atop models she doesn’t control and clones competitors ship in weeks.
Strip the messenger framing and Wabi’s value chain is a thin layer between the user and someone else’s frontier model — the layer that captures the least and is easiest to clone.
“Build me an AI therapy app.” Users never see code — the described app appears inside a shared chat.
Prompt orchestration + distribution + hosting. This is all Wabi owns — and users can rewrite its prompts.
ChatGPT or Gemini, chosen by the user in settings. The “intelligence” is OpenAI’s or Google’s, not Wabi’s.
A single-use artifact — a date-night planner, a trivia game. Novelty drives installs; habit is unproven.
No pricing, no margin over inference cost, no measured retention. The economics are entirely deferred.
Wabi’s suppliers are also its most dangerous competitors. OpenAI (ChatGPT Apps / GPT Store) and Google are building native “describe-an-app” capability; Meta (“Muse”) and OpenAI (“Dots”) are building shared agentic messengers directly. Wabi has no model moat, no pricing power over inference, and margin exposure to whatever its vendors charge.
Wabi’s $1.25B sits below app-builder leaders Lovable (~$13.3B) and Replit (~$9B) — companies with hundreds of millions in ARR — yet Wabi has $0 confirmed revenue. On a revenue basis it is the most expensive name in the set. The real anchor is the personal-agent mania: Instinct raised $1B at a $10B valuation the same week, still invite-only. Wabi is being priced in that wave, not against its own metrics.
Founder/CEO; built Replika. The valuation is substantially a bet on her track record and investor relationships
Lead investor; cites Kuyda’s “track record of great taste,” a qualitative, vision-based thesis
The PR headline — attached to zero published demand metrics; a free download is not a completed task
Replika’s lifetime users — a different product. Used as a halo stat; not evidence of Wabi engagement
Free today; pre-seed subsidizes usage. The hard question — who pays, at what margin over inference — is deferred
Material posted in a group reportedly “may remain visible” after you leave; account consent can’t cover a non-user’s data
Wabi’s valuation is lower than its app-builder rivals’ — but every one of them has real revenue, and Wabi has none.
Valued ~$13.3B (Aug 2026), with a reported ~$400M ARR. The highest-valued app-builder — and it monetizes. (Aggregator figures; treat as estimates.)
$400M Series D at a $9B post-money (Mar 2026); ARR disputed ($265M–$525M). Note: Replit CEO Amjad Masad is a Wabi angel — coopetition.
Bolt.new last ~$700M (Jan 2025); v0 rides Vercel (~$9.3B); Base44 was acquired by Wix. The “describe-an-app” feature is table stakes shipped on a weeks-long cadence.
The real valuation anchor: a personal-agent raising $1B at $10B the same week, still invite-only. Wabi is priced in this bubble, not the app-builder one.
No lawsuit or active controversy attaches to Wabi yet. But the privacy fault line is specific and precedented: Wabi ingests calendar, email, location, and Health data and surfaces agent output inside group chats — the exact data-sensitivity pattern that got Replika banned in Italy (2023) and reportedly fined €5M (2025), under the same founder.
Seven structural risks — several never addressed in the round’s coverage.
$1.25B on no monetization model and no disclosed usage metric. The multiple cannot be computed because revenue is $0. The price is reputation plus category hype — the froth signal in a single line.
No owned model; the product routes to ChatGPT or Gemini, whose owners ship the same capability natively. No inference-cost control, no pricing power, no moat — the suppliers are the competitors.
Mini-apps for one-off occasions are novelty artifacts. The category’s unsolved problem is retention, and there is no published cohort of active, repeat Wabi groups — installs are not habit.
Calendar/email/location/Health ingestion surfaced in group chats, with a group-data persistence clause and no mechanism to consent a non-user whose emails or calendar appear. A direct echo of Replika’s Italian ban/fine lineage — same founder.
Lovable, Replit, Bolt, and v0 ship the same feature on a weeks-long cadence; Meta and OpenAI are building shared/agentic messengers directly. The core feature set is commoditized.
“Free, model TBD” means the hardest question — will anyone pay, and at what margin over inference cost — is entirely unanswered, with usage subsidized by venture dollars.
Kuyda’s prior company monetized emotional attachment, abruptly removed features, and faced regulatory action and backlash. The “subsidize free usage, monetize later” playbook is identical — a deferred reckoning, not a resolved question.
Wabi’s round facts are rock-solid; everything that would justify the price is missing. This is a $1.25B price on $0 revenue, no disclosed usage, no owned model, and a feature competitors clone in weeks. The real explainer is the personal-agent bubble — and against the app-builder leaders that do have revenue, Wabi is, per dollar earned, the most expensive name in the field.
Based entirely on publicly available information, including the TechCrunch announcement of October 29, 2026. Figures are labeled by confidence; the round facts are well-corroborated while every operating metric is absent.
We build and maintain websites, help businesses of every size put AI to work, and run growth marketing that’s held to the same standard as our research.
