the-algorithm · Kashmiri Saffron · Live Shopify Store
From Context.md to Checkout
Every image below is a real artifact — rejected logos, hallucinated labels, and all. Nothing staged.
The 30-Second Version
Here's the whole thing, fast.
But the research came first.
Two rejected logo rounds. We weren't sure the third would land either.
Meet Michael
Michael Jagdeo
Operator, Exponent Labs
Rejected two full logo rounds outright rather than accept "close enough," and pushed for an actual diagnosis instead of a third guess — the fix turned out to be fine linework instead of flat fills, found by rasterizing the actual output and looking at it directly instead of re-guessing.
The Full Story
Every image on this page is a real screenshot from the actual workflow.
Phase 1
A father's trip to Kashmir was in the vault a month before Saffuja had a name.
07-03: logged in a MythosHealth strategy session as "undeveloped, no research." 07-04: Saffuja gets its name, split off as its own brand. 07-22: the supplement's manufacturer, Himalaya, closes as an option — no replacement found since.
Phase 2
It started with one file: context.md. Then the master prompt.
“Which form factors we should test first with a severely limited inventory of 1 kg and a $20/day Demand budget.”
— master-prompt-the-algorithm-saffron.md
Phase 3
The-algorithm didn't guess. It read 13 real files before we made one decision.
13
markdown files produced
3,250
lines of raw research
28
named competitors found
7
platforms searched
The real decision this research drove: 5 form factors (Whole Threads, Saffron Honey, Kesar Kahwa, Saffron Skincare, Saffron Encapsulated), and a price inside the $8–12/g gap between the ~$0.95/g wholesale floor and Diaspora Co.'s $22/g ceiling — almost nobody else was there.
Phase 4
"This is the colour and vibe we want to use."
“This is the colour and vibe we want to use.”
— Michael, on the reference photo
Phase 5
Three marks. All rejected.
“Those logos just have gold and grey. fail.”
— Michael, on round 1
Phase 6
Two colors now. Still rejected.
“Those don't look like luxury at all. Fail.”
— Michael — four words, no hedging
Phase 7
The problem wasn't color. It was flat fills instead of fine linework.
Rasterized the actual output and looked at it directly instead of re-guessing. Fine Radial Mandala — approved.
Phase 8
I — Whole Threads. The anchor SKU.
The most common single-unit size across every competitor studied (1g). Real checkout, Named-Farm angle.
Phase 9
II — Saffron-Infused Honey. Same kilogram, no new vendor.
Validated by two independent competitors' own cross-sell data. Producible from the same kilogram.
Phase 10
III — Kesar Kahwa. Kashmiril's #1 cross-sell.
5 stars, 34 reviews on Kashmiril's own listing. Same logic: same kilogram, no new sourcing relationship.
Phase 11
IV — Saffron Skincare. Coming Soon.
Real category signal — 2 competitors independently sell it — but needs a cosmetic formulator Saffuja doesn't have yet.
Phase 12
V — Saffron, Encapsulated. Coming Soon. One honest bug.
Its own fully-researched track — blocked on the same Himalaya manufacturer collapse logged in the pre-story.
“Supplement bottles kept hallucinating fake brand text nobody asked for — "Elixir Vitae," then "Aurum Vita," then a garbled "Supplement Si Muls." None of it was Saffuja. Four images got regenerated with an explicit blank-label instruction before this page ever saw them.”
— Build note
Phase 13
And then, a real store.
Everything above, actually live as a Shopify storefront — real products, real checkout, password saffron.
At a Glance
The whole workflow, by the numbers.
13
Real research files read
3
Logo rounds before approval
25
Real product images generated
5
Real product videos generated
725
Higgsfield credits — every image and video on this page
4
Images regenerated after AI hallucinated fake brand text
$600
Committed by Brian Ahuja to launch the paid campaign
Unlike Umami Salt, this one didn't run on a $-per-tool ledger — the real spend was 725 Higgsfield credits, plus a $600 launch commitment.
Meet the Investor
Brian Ahuja
MD · Neurologist · Investor
The thread underneath the deal: Kashmiri saffron access via his own father. Also Umami Salt's first investor — he watched that same sketch-to-brand process happen live in a group chat before writing a check.
This time, he's putting $600 behind the paid campaign to launch it — the result of watching this whole build happen.
The Investment
Brian Ahuja was Umami Salt's first investor too — he watched that same sketch-to-brand process happen live in a group chat before writing a check.
This time, he's committing $600 to launch Saffuja's paid campaign — the result of watching this whole build happen, on top of the fact that his own father's Kashmir saffron access is the real thread underneath the deal.
This is what it looks like when the process stays visible — rejections included.
Context.md → Research → Reference → 3 Logo Rounds → 5 Real Product Lines → A Real Store. Built by Exponent Labs — the same syndicate model behind every case study on this page.