Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Shann³ Profile picture
Shann³
@shannholmberg
I cover AI marketing & growth. Sharing every framework as I build it. Founder @espressioai, @lunarstrategy
Apr 29 • 10 tweets • 3 min read
how to get AI SEO articles indexed and ranking in under 14 days

most SEO workflows break between keyword research, drafting, review, and distribution

at Espressio we turned ours into one operating loop, 7 agents passing work from keyword backlog to distributing backlinks

what kills the slop is a human input that system waits on before anything gets written

here´s how it runs 🧵Image 2/9 SERP agent

pulls live US and UK results, competitor pages, related questions, our own source material, existing site inventory, GSC and GA4 signals

reads what already ranks. listicles, how-tos, comparisons, templates, service pages

output is the beat plan, what the article has to cover, the angle that makes it more useful, where we can add workflow detail or founder POV
Apr 20 • 9 tweets • 3 min read
how OpenClaw compares to Hermes Agent

Elvis spent 9 hours reading the source code of both side by side, comparing how they handle skills, self-improvement, and governance.

here´s everything you need to know 🧵 Image hermes by @NousResearch self-improves by writing its own skills.

the system prompt nudges the agent: every N tool calls, consider saving a skill. after task completion, a background review scans for skill-worthy patterns. before context compression kicks in, durable knowledge gets flushed to disk.

Elvis watched it create an extract-social-testimonial skill on its own, something he never would have thought to build manually.
Apr 12 • 7 tweets • 2 min read
how to build your own content engine

Ronin runs 10 social accounts without writing a single post, no content team, just 17 markdown files and one AI agent

here's how it works 🧵 Image the problem:

most people open claude, type "write me a post about X", get generic output, spend 20 minutes fixing it, then repeat for every platform

the fix: give the AI a folder of .md files with your voice, platform rules, hooks, and audience profiles before it writes anything
Apr 5 • 10 tweets • 3 min read
karpathys LLM obsidian knowledge base went completely viral this weekend

I found 7 important things that separate a wiki you keep from one you abandon

save this and send it to your agent (prompt in last tweet)🧵 Image quick recap of what karpathy built:

instead of uploading files and asking questions every time (RAG), the LLM reads your sources once and compiles a structured wiki. summaries, cross-references, contradictions, all maintained by the LLM.

you curate sources and ask questions. the LLM handles the rest.

raw sources → wiki → schema
Mar 31 • 9 tweets • 3 min read
how to stop claude from being a YES-MAN

Ole built a skill that forces 5 AI advisors to argue about your question, blind-review each other, and hand you a verdict you can trust

Here's how it works and how to set it up 🧵 Image the problem: you ask claude "should I launch this product?" and it finds 5 reasons why you should

then you ask "is this product a bad idea?" and it finds 5 reasons why it is

same product with different framing and completely opposite answers

your assumptions shaped the response and claude just agreed with you
Mar 24 • 8 tweets • 2 min read
Karpathy's AutoResearch is changing how campaigns get optimized and most marketers haven´t heard of it yet.

Ole Lehmann tested it on landing page copy, 56% → 92% pass rate overnight.

here´s how it works for marketing / skills 🧵 Image 2/8 the pattern is called AutoResearch (Karpathy coined it for ML but it maps to marketing perfectly)

how it works:

> AI edits one variable in your copy
> runs a short test against a single metric
> measures the result
> keeps wins, discards losses
> repeats

~12 iterations per hour. while you sleep
Mar 18 • 7 tweets • 2 min read
a claude code engineer at anthropic just published every skill type worth building

9 categories, the exact mistakes that waste your time, and how they structure hundreds of skills internally

here´s what you need to know 🧵 everyone thinks a skill is a markdown file you drop in .claude/

Wrong, a skill is a folder, with scripts, assets, data and templates

stuff the agent can discover and run, not just read
Mar 14 • 8 tweets • 2 min read
claude superpowers is the most underrated plugin for marketers right now

83,000 github stars. trending daily. but almost everyone using it is a developer

here´s how it works and how to apply it to marketing 🧵 Image superpowers is a skill framework for claude code

it forces a structured pipeline before any work gets done

> brainstorm first, ask clarifying questions
> lock in a spec before executing
> break work into small tasks
> execute with subagents in parallel
> review everything before shipping

no more "just start writing." think first, then build
Oct 2, 2022 • 15 tweets • 2 min read
Brand communities have been around for decades

But they’re now a more important part of the marketing mix than ever before

(Thread) As Web3 is driving community building into the spotlight, many traditionally web2 companies are trying to jump on this bandwagon