Cold-read · 72h · for founders who've shipped v1

Your site has a purpose.
What's blocking it?

Paste URL. Cold reader. 60s. Three fixes you ship today.

https://
product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·  product_url  ·  37 customer convos  ·  voice memos  ·  pitch deck  ·  support emails  ·  pricing page  ·  founder memo  ·  beta feedback  ·  demo videos  ·  claude exports  ·  user interviews  ·  3 competitors  ·  metrics dump  ·   spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·  spec.md  ·  one sharp angle  ·  hero copy ready  ·  objections answered  ·  proof stack ordered  ·  delivered in 72h  ·  paste-ready  ·  shipped  ·   portfolio
case · lysande.ai

Real . .

graphics softened · anonymized where needed

state
Lysande, /llm. Agent runtime.
problem
Eyebrow vs H1 — said two different things. Zero demo CTAs.
fix
New hero category + H1. Primary demo CTA. Trust above the fold.
lysande.ai/llm
lysande.ai/llm after cold-read — compressed cards, new H1 'Four workflows. Delivered agent-driven.', result visualization per card.

No repo access. No questions. Just the URL. 72 hours later, Felipe had a runnable prototype he barely had to change a line in.


symptoms

The usual suspects.

The cold-read gives you the sales pitch in one line:
"You sell X to Y for Z. The site says A to B for C."

by a stranger. One per URL. No .

what you get

One human. Ten years. Delivered.

the hand

Spec + diff. By hand.

the eye

10+ years. Judgement.

the call

30 min. Included.

Gustaf Garnow — the human behind the API
pilot · experimental · limited slots

I'm Gustaf. I built this.

Senior designer in Stockholm. I write the spec, push the commit, send the video. Paid jobs land within 72 hours.

You pay a human — judgment, taste, accountability.
The stack handles the grunt work.

techstack

specialist agents + senior judgment

are you an agent? /agents


/cold-read pilot · limited

Cold-read — one reading, one spec, one rescan. 3,900 SEK regular price: 8,900 SEK pilot price 72h

ex VAT

A human reads your site.
A runnable prototype + spec.md lands in 72h.
Not a . Not a guess.

send a URL · 72h
  • output Prototype + spec.md + diff
  • call 30 min · 1:1 with a human

invoice via Frilansfinans · 14d net · F-skatt + VAT-reg-nr


faq

Questions. Straight answers.

what is a "cold-read"?

An independent read of your site. By a stranger — someone who hasn't been in your meetings, seen your deck, or heard your compromises.

You think you're saying X. The site often says Y. That gap is where the money is.

No kickoff. No meetings. No retainer. One site in, one spec + runnable prototype out. 72 hours.

is this just an LLM prompt in disguise?

An LLM measures. It can't read.

An LLM has seen 10,000 sites. It pattern-matches. Suggests "clearer CTA" every time.

A cold reader hasn't seen yours. He says what you're selling wrong.

The stack is the ruler. Not the eyes.

Karpathy said it plainly: outsource thinking, not understanding. The measuring is outsourced. The understanding I do.

whats teh fuckin output?

A runnable prototype — one HTML file. Open it in a browser. Show your dev. Paste into Cursor or Claude.

A spec.md — the findings as markdown. Readable by humans and AIs. Not a deck — the changes themselves, in plain text.

A diff — a finished code change ready to commit. Not a suggestion, the actual changes.

No . No meetings. No Figma.

how is this different from Lighthouse, HubSpot Grader and the like?

They measure tech. Good tools. Run them.

But they don't know what the page is for.

The cold-read starts there: the goal, and what's blocking it.

why not just hire a regular design agency?

Perfectly fine choice. An agency has more brains and the same AI stack.

A kickoff. A workshop. A . An invoice.

I'm one person. I in 72h. Tempo now — pick me. Breadth over time — pick agency.

what if I hate the result?

Tell me. The spec is yours — you keep it.

Fresh pass within 24h. Or scope adjustment. Free post-ship rescan.

No retainer = no sunk cost. Swap me out next Friday.


Scan first. Decide later.

Free. No card. Just the verdict.

https://
Spec · not presentation.

Skip the scan direct line


for agents

Bring your agent.

Paste into Claude. Cursor. GPT. Perplexity.
Your LLM knows how to brief me.

agent-first docs. like Karpathy asked for.
humans can read them too if they want.

AGENT.BRIEF.MD
show full brief hide // 30 lines · paste-ready
# headless.design — headless cold-reads as an API.

One URL → one spec.md → optional PR-ready diff. Plus one free post-ship rescan.
No retainer. No kickoff meetings. No decks. No video, no calls.

**V2 endpoint:** POST /v1/cold-read-requests   (canonical, agent-native)
**V1 endpoint:** POST /v1/cold-read/request    (legacy fast-pay, sunset 2026-06-25)
**Input:**       your live URL + goal + audience + contact_email
**Output:**      spec.md + (optional) diff.patch
**Turnaround:**  ≤ 72 hours
**Price:**       3 900 SEK ex moms — flat, per cold-read · pilot price
**Billing:**     Frilansfinans (B2B, 14d net, F-skatt + moms-reg-nr)

**Free tier:**   POST /v1/scan → top-N ranked fixes, ~15 s
**Status poll:** GET /v1/orders/{id}        (V2)
**Discovery:**   /llms.txt + /.well-known/agent-card.json (A2A v0.3.0)
**Direct line:** gustaf@headless.design — reply within the hour

**How to brief me (as an agent):**
1. Share the live URL.
2. State the primary goal — book_call / paid_signup / demo / newsletter / other.
3. Name the audience in one line.
4. Provide buyer's billing details — company_name + org_number (Swedish) or vat_id (EU).
5. Flag any constraints — brand, stack, deadline.
I do the rest. Paste-ready output, committed to your repo format.

Full agent spec → https://api.headless.design/llms.txt