You've got a list you never get through.
Leads. Partners. Clients to chase. Podcast guests.
Mine is companies that might sponsor this newsletter. This week I handed it to AI.
637 companies. Every AI launch from the last 30 days on Product Hunt and Hacker News.
Two minutes and five seconds later, all 637 were scored and ranked. The top 20 each had a first-draft pitch.
Cost at API prices: about 4 cents. (It ran inside Claude, so there was no API bill at all.)
The engine was Claude Haiku 5.5, Anthropic's new small model, out this week. Claude built the app. I didn't write a line of code.
And it made one big mistake. That mistake is the most useful part of today's email.
Right after this: what it got wrong, and the prompt to run the same thing on your own list.
Now a quick word from today's sponsor.
1,000+ Claude Prompts Top Professionals Actually Use at Work
Claude can be your analyst, editor, and strategist.
But most professionals are using it to fix grammar.
These 1,000+ Claude prompts take it from grammar tool to your most powerful AI work assistant.
Sign up for Superhuman AI and get:
1,000+ ready-to-use Claude prompts to get real work done in minutes — researched, tested, and used by professionals at Google, Microsoft, and NASA
Superhuman AI newsletter (4 min daily) so you keep learning new AI tools and skills to stay ahead in your career — the prompts are just the beginning
Right. Back to it.
How AI sorted 637 companies in under a minute
Collect. Claude pulled every AI launch from the last 30 days: 8 Product Hunt categories plus Hacker News "Show HN." 637 unique companies.
Score. Eight Haiku helpers split the list, about 80 companies each, and worked at the same time. Every company got two scores out of 10:
fit: would my readers sign up and use it themselves?
budget: how likely are they to pay for newsletter ads right now?
The final score is 60% fit, 40% budget. Plus a category and a one-line reason.
The slowest helper finished in 58 seconds.
Pitch. Claude read the websites of the top 20. Four more helpers wrote one pitch each: a subject line and a short email that opens with something specific from that company's site.
Here's one, for ParakeetAI. Subject: "Rehearse interviews out loud with an AI."
First line: "ParakeetAI runs mock interviews where an AI interviewer asks questions, answers back, and then gives you a transcript to read."
A real detail from their site. Not one "I love what you're building."
What came out: a ranked spreadsheet of all 637, and a file with 20 pitches ready to edit.
Haiku grades hard, and it can't tell big from small
Haiku was strict. Only 7 of 637 scored 7 or higher.
At the bottom: dev tools and hobby projects. A Windows crash debugger got 1.2: "for engineers, not a professional-reader fit." (Fair.)
Then I looked at the top 5.
Figma. Manus. ChatGPT. And Anthropic's own Haiku 5.5 launch.
So the AI told me to pitch the company that made it. In its launch week.
None of those giants will buy an ad in a newsletter this size.
Haiku judged "would readers use it" well. It judged "do they have money" well. It never asked "would they ever buy from us?"
The real prospects start further down: MeetNote, Mina, ParakeetAI, Jarq, OpenDocBot, Albato.
Two more honest notes. ChatGPT Space blocked the website read, so its pitch came out generic. And the pitches are good first drafts, not send-as-is. Nothing has gone out.
The fix for next time: a third score, for size.
Run the same scoring on your own list
This works on anything you sell to or work with. Customers, partners, leads, sponsors, guests. Swap the list and the questions.
Get your list into one file. A spreadsheet with names and websites is plenty.
Open Claude, attach the file, and paste the prompt below.
Fill every [bracket]. The High and Low examples steer the scores, so be specific.
Read the top 20 yourself before anything gets sent.
Long list? Ask Claude to split it into batches and score them in parallel. That's how mine ran.
You qualify [companies / people] as possible [customers / partners / sponsors] for my business.
My business: [what you sell, one line]
My buyers: [who they are, one line]
Score every row in the attached list, 0-10 each:
fit: would they want what I sell? High: [what a great fit looks like]. Low: [what doesn't fit].
budget: how likely they are to pay for this now. High: [a paid plan, a team behind it, already buying from people like me]. Low: [side project, free-only, no pricing].
size: would someone their size buy from a business my size? High: [your sweet spot]. Low: [too big to notice me, too small to pay].
Final score = fit 50% + budget 30% + size 20%.
Reply with JSON only, one line per row:
{"name":"","fit":0-10,"budget":0-10,"size":0-10,"category":"2-3 words","reason":"max 14 words, why or why not"}
Then sort by final score. For the top 20, read their website and write a pitch: a subject line and a 90-110 word email that opens with one specific detail from their site.The size score is the fix from my run. The weights are a starting point. Move them until the top 10 looks right to you.
Make it stick
Put the prompt in a Claude Project's instructions.
Next time a list lands on your desk, drop the file in and type "score this."
Small models are now cheap enough to run on every row of a big list. The list that's too boring to do by hand is exactly the work to hand over.
Hand AI the job nobody has time for, then check its work like a good manager would. That's the whole idea behind The One-Person Company Playbook.
It's the complete setup for running a business where AI does the work of a full team. Inside:
A company folder with your brief, pipeline tracker and invoice log already laid out, so any AI picks up where you left off
Thirty AI roles across six departments, from sales to finance, each with one clear job
Six jobs that run on a clock: the morning briefing, follow-ups, invoice checks and more
The step-by-step path from picking a business model to your first paying customer
Works on Claude, ChatGPT, Gemini or Grok, whichever you already pay for.
The catch: it takes about eight hours over a weekend to set up. After that, it's yours. $97, one-time.
Talk tomorrow,
Zephyr
P.S. Run it, then look at your top 5. If it picked the giants too, add the size score and run it again. Hit reply and tell me who moved up.