Pricing-page honesty: disclose provider usage cost before checkout (two-bill story P0) #163

2026-07-22T22:30:26Z

tim.lingo commented

2026-07-22 22:30:26 +00:00

Root cause

A paying customer could reach checkout without ever learning they will also pay an AI provider. Every tier is bring-your-own-key until Neuron Inference launches, but the pricing page never said so before money moved — and the hero's "Priced below ChatGPT on day one" claim invited a total-cost misreading (Neuron subscription + provider usage can exceed a ChatGPT subscription). This is the OpenCat failure mode: the 1-star reviews aren't about the price, they're about finding out about the second bill after paying the first.

The change (copy only — no code-behavior changes)

src/hero.el — hero sub-line now reads "Runs on your machine. Remembers everything. The memory is yours — forever." A bring-your-own-key user's total cost isn't ours to promise; the price claim is replaced with a claim we own outright.
src/pricing.el — a one-line price rider under each tier price, in the cards' existing secondary-line style:
- Free: "$0 + your AI provider's usage — free provider option available"
- Professional: "+ your AI provider's usage (typically $2–8/mo for everyday chat) · one bill once Neuron Inference launches"
- Founding Member: "+ your AI provider's usage until Neuron Inference launches — then it's included, forever"
src/pricing.el — a "What you'll need" box above the tier cards (reuses the existing founding-banner box pattern) explaining the provider connection, the free Gemini option and its training caveat, typical costs, that Neuron adds no markup, and the one-bill future under Neuron Inference.

How to test

Visual review of the hero and pricing sections: hero sub-line reads the new claim; each of the three tier cards shows its rider directly under the price; the "What you'll need" box renders between the pricing header and the tier grid. Both edited files compile clean through elc (exit 0) and the new strings are present in the emitted C.

Cost-figure provenance

The $2–8/mo figure was computed 2026-07-22 from provider prices fetched from official pricing pages the same day (per-million-token input/output): Anthropic Sonnet $3/$15 intro $2/$10; OpenAI gpt-5.5 $5/$30, nano $0.20/$1.25; Gemini Flash $1.50/$7.50, Flash-Lite $0.10/$0.40; xAI Grok 4.5 $2/$6 — under the assumption of ~15 msgs/day × (1,500 in + 400 out) tokens.

🤖 Generated with Claude Code

## Root cause A paying customer could reach checkout without ever learning they will also pay an AI provider. Every tier is bring-your-own-key until Neuron Inference launches, but the pricing page never said so before money moved — and the hero's "Priced below ChatGPT on day one" claim invited a total-cost misreading (Neuron subscription + provider usage can exceed a ChatGPT subscription). This is the OpenCat failure mode: the 1-star reviews aren't about the price, they're about finding out about the second bill after paying the first. ## The change (copy only — no code-behavior changes) 1. **`src/hero.el`** — hero sub-line now reads "Runs on your machine. Remembers everything. The memory is yours — forever." A bring-your-own-key user's total cost isn't ours to promise; the price claim is replaced with a claim we own outright. 2. **`src/pricing.el`** — a one-line price rider under each tier price, in the cards' existing secondary-line style: - Free: "$0 + your AI provider's usage — free provider option available" - Professional: "+ your AI provider's usage (typically $2–8/mo for everyday chat) · one bill once Neuron Inference launches" - Founding Member: "+ your AI provider's usage until Neuron Inference launches — then it's included, forever" 3. **`src/pricing.el`** — a "What you'll need" box above the tier cards (reuses the existing `founding-banner` box pattern) explaining the provider connection, the free Gemini option and its training caveat, typical costs, that Neuron adds no markup, and the one-bill future under Neuron Inference. ## How to test Visual review of the hero and pricing sections: hero sub-line reads the new claim; each of the three tier cards shows its rider directly under the price; the "What you'll need" box renders between the pricing header and the tier grid. Both edited files compile clean through `elc` (exit 0) and the new strings are present in the emitted C. ## Cost-figure provenance The $2–8/mo figure was computed 2026-07-22 from provider prices fetched from official pricing pages the same day (per-million-token input/output): Anthropic Sonnet $3/$15 intro $2/$10; OpenAI gpt-5.5 $5/$30, nano $0.20/$1.25; Gemini Flash $1.50/$7.50, Flash-Lite $0.10/$0.40; xAI Grok 4.5 $2/$6 — under the assumption of ~15 msgs/day × (1,500 in + 400 out) tokens. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

tim.lingo added 1 commit 2026-07-22 22:30:27 +00:00

copy(pricing+hero): disclose the two-bill reality before money moves — price riders, What-you'll-need box, hero claim fix c38a552266

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

tim.lingo commented

2026-07-22 23:04:36 +00:00

Counsel flag (2026-07-22 terms/privacy audit): the What-you'll-need box asserts a named third party's training practice (Google free tier trains on chats / paid tier doesn't). Grounded same-day in Google's own pricing page, but this claim class goes to attorney review with a dated source before deploy — review rides the held #160/#161 batch. Details in neuron-ui#193.

will.anderson reviewed 2026-07-28 03:17:43 +00:00

will.anderson left a comment

Review: honesty-before-comfort pass. Content-readiness only — the website is under the deploy-hold pending counsel, so this is feedback for when Daniel clears it, not a merge/deploy signal.

Verdict: the disclosure content is accurate and clear — approve the content, with one completeness gap to close before it ships.

What's right:

Hero: dropping "Priced below ChatGPT on day one" is the correct call — it invited a total-cost misread (Neuron subscription + provider usage can exceed a ChatGPT subscription). I scanned src/ for other total-cost claims: the remaining "priced below the major APIs" lines are all narrower Neuron-Inference token-pricing claims — defensible, not total-cost — so the hero was the one real offender and you got it. The replacement claim is one we own outright.
Per-tier riders are honest and specific; the $2–8/mo figure is grounded and consistent between the Professional rider and the "What you'll need" box.
"What you'll need" box is genuinely complete: provider connection, the Gemini free-tier training caveat (and that the paid tier doesn't train — more complete than the app copy), "Neuron adds no markup and never touches the bill", and the crucial "Heavy agentic use costs more... you're always billed at your provider's own rates." This is honesty-before-comfort done right.

One gap — the disclosure stops at the pricing page, but this PR's own root cause names checkout:

src/checkout.el is the literal money-moves step and still says nothing about the provider bill. "Bring your own API keys" appears as a feature bullet, but there is no "+ your AI provider's usage" rider and no two-bill note anywhere on the page. A user can click (or deep-link) straight to checkout and pay the first bill without ever seeing the pricing-page box. Recommend carrying the one-line rider onto checkout.el — plus a compressed "you'll also connect a provider that bills you directly" note — before this goes live. That checkout screen is exactly where the OpenCat 1-star "second bill" failure happens.

One adjacent, pre-existing flag (not introduced by this PR, but relevant to billing-model honesty since the review asked): the Free-tier feature list on this page implies more than the shipped free scope (chat + markdown + memory) — it lists "Unlimited projects", "Neuron Inference included when it launches", and "Core built-in capabilities". If the paid $19 / $199 tiers are what unlock the agentic layer, the Free card currently reads richer than it is, and "Neuron Inference included" on Free sits oddly beside the Founding "then it's included, forever" perk. Worth reconciling with the real tier boundary before the page ships — separate from this copy PR.

Net: strong, honest disclosure. Close the checkout gap and it's ready for Daniel's clearance.

Review: honesty-before-comfort pass. **Content-readiness only** — the website is under the deploy-hold pending counsel, so this is feedback for when Daniel clears it, not a merge/deploy signal. **Verdict: the disclosure content is accurate and clear — approve the content, with one completeness gap to close before it ships.** What's right: - **Hero**: dropping "Priced below ChatGPT on day one" is the correct call — it invited a total-cost misread (Neuron subscription + provider usage can exceed a ChatGPT subscription). I scanned `src/` for other total-cost claims: the remaining "priced below the major APIs" lines are all narrower Neuron-Inference *token-pricing* claims — defensible, not total-cost — so the hero was the one real offender and you got it. The replacement claim is one we own outright. - **Per-tier riders** are honest and specific; the $2–8/mo figure is grounded and consistent between the Professional rider and the "What you'll need" box. - **"What you'll need" box** is genuinely complete: provider connection, the Gemini free-tier training caveat (and that the *paid* tier doesn't train — more complete than the app copy), "Neuron adds no markup and never touches the bill", and the crucial "Heavy agentic use costs more... you're always billed at your provider's own rates." This is honesty-before-comfort done right. **One gap — the disclosure stops at the pricing page, but this PR's own root cause names checkout:** - `src/checkout.el` is the literal money-moves step and still says nothing about the provider bill. "Bring your own API keys" appears as a feature bullet, but there is no "+ your AI provider's usage" rider and no two-bill note anywhere on the page. A user can click (or deep-link) straight to checkout and pay the first bill without ever seeing the pricing-page box. Recommend carrying the one-line rider onto `checkout.el` — plus a compressed "you'll also connect a provider that bills you directly" note — before this goes live. That checkout screen is exactly where the OpenCat 1-star "second bill" failure happens. **One adjacent, pre-existing flag** (not introduced by this PR, but relevant to billing-model honesty since the review asked): the Free-tier feature list on this page implies more than the shipped free scope (chat + markdown + memory) — it lists "Unlimited projects", "Neuron Inference included when it launches", and "Core built-in capabilities". If the paid $19 / $199 tiers are what unlock the agentic layer, the Free card currently reads richer than it is, and "Neuron Inference included" on Free sits oddly beside the Founding "then it's included, forever" perk. Worth reconciling with the real tier boundary before the page ships — separate from this copy PR. Net: strong, honest disclosure. Close the checkout gap and it's ready for Daniel's clearance.

You are not authorized to merge this pull request.

This pull request can be merged automatically.

View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u origin copy/two-bill-story:copy/two-bill-story

git checkout copy/two-bill-story

Sign in to join this conversation.

2 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: neuron-technologies/neuron-web#163