well roast

**opus 4.8 thinking roast: **

I've got both. Here's the roast, straight up — measured against your own Product Hunt playbook.

The premise is solid, the framing is a mess

The good news first, so the burns land harder: Well actually clears your hardest bar. "Robots chasing invoices & receipts" via an AI-native infrastructure that turns any web portal into an API is a genuine pain point with a defensible moat — connectors built and self-healed by AI across 10,000+ providers is not a weekend clone . Your playbook says build where "deep expertise, unique data, proprietary technology" protects you from saturation . Max ran spend management at European SME banks; he checks that box. So why does the launch read like it's apologizing for itself?

Roast #1: the tagline hooks, the pitch un-hooks

Your rule: "crystal-clear value proposition that hooks in 5 seconds" . "Streamline payables with robots chasing invoices & receipts" is genuinely great — punchy, visual, instant . Then the maker comment immediately drowns it in "embeddable infrastructure for Spend Management and Bookkeeping tools," "white-glove onboarding," and "360° connector infrastructure" . You went from robots chasing receipts to enterprise B2B2B middleware deck in two scrolls. The 5-second hook is real; the 30-second follow-through fumbles it.

Roast #2: you launched the thing that isn't the product

This is the cardinal sin from your own list — "Launch unfinished or waitlisted products... closed-betas" . Read the copy: "Book a call with us to get started," "Coming soon: Well App 1.0," "we may have a mass market product we will release in september" . The Product Hunt crowd — makers and founders — are the people who feel the receipt-chasing pain, and you've explicitly told them their product is "coming soon" while pitching them a B2B integration that requires a sales call . You launched the API for your competitors instead of the app for the audience in the room. That's a "book a call" CTA on a community platform that rewards "try it now."

Roast #3: "show, don't tell" became "tell, then book a call"

Your playbook: "Don't just claim you're 'better' – show exactly HOW" with a demo and first screenshot . The single best line in the whole thread came from a commenter, not the maker: "This was the shortest video explainer I have seen, and it told so much" . So the demo apparently slaps — but the maker buried it under a wall of bullet-pointed feature lists (native drivers, mobile channels, Chrome extension, privacy architecture, flexible output) instead of leading with the thing that made a stranger say "wow" . You had the show. You chose to tell.

Roast #4: the self-inflicted credibility wobble

First top comment is literally "wonder how reliable the AI extraction really is" . The answer — "self-healing... AI is replacing thousands of engineers" — is a bold flex that raises the reliability question instead of closing it. Your playbook warns against launching with anything that smells "buggy" or "unfinished" ; leading with "robots replacing thousands of engineers" without a hard accuracy number invites exactly the skepticism the first commenter voiced. Confidence is good; an unbacked superlative on your #1 doubt is a layup for critics.

Scorecard against your own rubric

Playbook criterion	Verdict on Well Embed
Solves a clear pain originally	Pass — receipt/invoice chasing is real and the API angle is fresh
Defensible niche / hard to clone	Strong pass — self-healing AI connectors, ex-bank expertise
5-second value prop	Half pass — killer tagline, then buried in jargon
Show, don't tell	Fail — feature dumps over the praised demo
Don't launch waitlisted/coming-soon	Fail — "App 1.0 coming soon," "book a call" CTA
Differentiate loudly	Half pass — moat exists but never stated as a number or hook

The one-line fix

If you wanted to relaunch this the way your own article preaches: lead with the demo video that wowed people, put one hard extraction-accuracy stat in the first comment, kill the "book a call" CTA in favor of a live self-serve playground, and stop apologizing that the consumer app is "coming soon" — either launch that to founders or commit fully to "the invisible receipt-fetching layer your accounting tool should've had." Right now it's a defensible product wearing a hesitant, half-finished launch.

Want me to draft a rewritten maker's first comment + tagline that fixes all four roasts?

The Tagline Faceplant

"Streamline payables with robots chasing invoices & receipts" — your rubric demands a value prop that "hooks in 5 seconds," and instead you opened with robots cosplaying as collection agents . The word "payables" is accountant-speak that immediately repels the founders you claim to be saving, and "robots chasing" sounds less like infrastructure and more like a Roomba late on rent . Show, don't tell — and "robots chasing" tells me nothing about how .

Positioning: Pick a Lane, Maxime

Your own checklist's #1 sin is vague product positioning, and this launch commits it enthusiastically . Is it Well Embed? Well Extract? Well App 1.0 "coming soon"? The maker comments toggle between three products and a "mass market product we will release in september" that nobody asked about yet . You're launching one thing while teasing two others — that's not a roadmap, that's positioning that hasn't decided what it wants to be when it grows up.

The Vaporware Tell

The rubric explicitly axes "waitlisted products, vaporware, closed-betas" — and your CTA is literally "Book a call with us" with "white-glove onboarding" . That's not a launch, that's a sales meeting wearing a launch costume. A Product Hunt audience wants to click and try; instead they get a calendar invite and a hint that the real product (Well App) ships in September .

"Self-Healing AI" — Bold Claim, Zero Demo

When a commenter politely asked how reliable the AI extraction actually is, your answer was that AI "is replacing thousands of engineers" and the connectors are "self-healing" . Your rubric says show, don't tell and don't just claim you're better — yet the reliability question got a buzzword sandwich instead of a metric, an accuracy rate, or a single screenshot of it recovering from a broken portal .

Defensibility: The Weekend Test

Your article's sharpest line: "If someone can rebuild your product in a weekend, rethink your strategy" . An "AI-powered Chrome extension that turns any web portal into an API" is exactly the kind of thing a hundred YC teams are shipping this quarter — the moat isn't the AI scraper, it's the 10,000+ connectors and bank-grade compliance, and you buried that under robot metaphors . Lead with the defensible niche (your SME-bank domain expertise), not the commoditized part .

What Actually Scored Points

Credit where due, per your own "Will Get You Featured" column :

Real pain, real credibility — building this at top-tier European SME banks is genuine domain expertise, the kind of moat your article praises .

The explainer video got unprompted praise for being short and information-dense — that's the "demo that tells the story fast" you preach .

Engaged maker — Max replied to every comment, which is the "community-driven" energy PH rewards .

The One-Line Verdict

By the gospel of your own Substack: solid founder, real pain, defensible domain knowledge — wrapped in vague positioning, a "book a call" CTA that smells like a closed beta, and a tagline that hid the actual moat behind a robot chasing receipts down a hallway .

Want me to rewrite the tagline and first comment so they'd actually pass your own checklist?

eonist/well roast.md

Select an option

No results found