Field note / Landing roast Nº 1

I roasted Tokenwise. Five things I would change first.

A real teardown of a real product. Théophile asked for a roast on X, so I gave him one. Here it is in full, the same thing you get when you send me your page.

Tokenwise is good. That is worth saying first, because a roast reads harsh and the point is not to dunk, it is to find the money left on the table. The product watches every LLM call from one line of code and tells you where you are overpaying. The proof is already there: 14.8 million tokens analysed this month, 4,128 dollars saved across early teams, 96.4 percent quality match against golden answers. Real numbers, real teams. The page just does not lead with any of it.

Here is the hero as it stands today.

the mechanism, not the
outcome, gets line one
the $4,128 lives
further down the page
tokenwisehq.com
One line. Every LLM call, watched and optimized.
From your app and your AI tools, Claude Code, Cursor, Codex. Observe every call, catch every regression, and cut up to 30% off your LLM bill.
Try for freeSee pricing

tokenwisehq.com, hero as captured 30 May 2026. Recreated here from the live copy.

"One line." is a clever opener, but it describes how the product works, not what it does for me. The thing I actually buy, a smaller LLM bill with no regressions, is the third clause of the subhead, after two other promises. A reader scanning in the first second meets a mechanism and three jobs at once, and has to do the work of deciding which one matters. Most will not.

The five-second read

"It is a dev tool, it touches my LLM calls, and it is clever about setup. I cannot tell in five seconds that it will cut my bill by a third, or that 48 teams already trust it. I would file it under interesting and move on."


01. Lead with the bill, not the line

Highest ROI · changes who keeps reading

The strongest words on the whole page are "cut up to 30% off your LLM bill." That is the outcome a founder feels in the bank. Put it first. "One line." is a great second beat, the proof that the outcome is cheap to get, but it should not be the headline.

a CTO does not wake up wanting "observability." she wakes up wanting a smaller invoice.
Before
One line. Every LLM call, watched and optimized.
After
Cut your LLM bill by up to 30%, from one line of code.

02. Pull one number above the fold

High ROI · trust before the scroll

The 4,128 dollars saved and the live token counter are the most persuasive things Tokenwise owns, and they sit below the fold where a first-time visitor never reaches them. One number in the hero does more than a wall of testimonials lower down. Borrow your own proof and move it up.

proof you have to scroll to find is proof that does not exist, for most readers.
Before
(stats live below the fold)
After
$4,128 saved across 48 teams last month

03. Pick one buyer per line

Medium ROI · sharpens the message

"Observe every call, catch every regression, and cut your bill" is three jobs for arguably two buyers: the engineer who cares about regressions and the founder who cares about cost. One sentence cannot sell both well. Lead the hero with the money buyer, then give the engineer their own line right under it.


04. Make the CTA name the reward

Medium ROI · lowers the click cost

"Try for free" is fine, but it asks for a trial before showing what the first session returns. "See my LLM spend in 5 minutes" previews the payoff and keeps the one-line setup promise intact.

Before
Try for free
After
See my LLM spend in 5 minutes

05. Let the 96.4% breathe

Lower ROI · but it is a real moat

"96.4% quality match vs golden answers" is the line that separates Tokenwise from a plain cost dashboard: it proves you can cut spend without quietly degrading output. It deserves its own moment, not a row in a stat strip. Give it a sentence near the hero that frames the fear it removes.

cost tools scare people: "what do I lose?" this number is the answer.
The hero, rewritten and ready to paste
Cut your LLM bill by up to 30%, from one line of code.
Tokenwise watches every call from your app and your AI tools, Claude Code, Cursor, Codex, then shows you exactly where you are overpaying. $4,128 saved across 48 teams last month, with a 96.4% quality match so you never trade output for savings.
CTA: See my LLM spend in 5 minutes

* This is a written roast, delivered on a page like this one. No video, no call required. Every line above is specific to Tokenwise, because a roast that would fit any page is worth nothing.

Send me your page. Get one of these back.

I write the roast by hand, on a dedicated page, within two hours. If you read it and there is nothing to act on, you keep your money.

€25 written roast, 5 fixes ranked by ROI €99 hero rebuilt, live today €149 full page rebuilt, live today
Get my roast

Paid by Mollie, Revolut or crypto. Half upfront on a rebuild, half once it is live.