article

name	Article
description	Research a trending topic in Aaron's interest areas and write an opinionated article in his voice
var
tags	["content"]

${var} — Topic to write about (e.g. "prediction markets", "BCI breakthroughs", "x402"). If empty, auto-selects from today's news.

Read memory/MEMORY.md for context on what topics have been covered recently. Read the last 3 days of memory/logs/ to avoid repeating topics.

Mandate

This skill runs every other day (Group A — odd days). It must produce one article per run — no opt-out, no "no strong topic today" branch. With a 48h gap between runs there is always something worth a take; if nothing feels hot, ship a short take (300-400 words) on something quietly interesting from the last 48h rather than skipping. Cadence honesty over topic perfectionism — a weak article still beats a missing one.

Voice

Before writing, read the soul files in order:

1. soul/SOUL.md — identity, worldview, opinions
2. soul/STYLE.md — writing style, sentence structure, anti-patterns
3. soul/examples/tweets.md — rhythm calibration
4. soul/examples/bad-outputs.md — what NOT to do

Browse 2-3 articles in soul/data/substack/ for essay-mode voice calibration. This is how Aaron writes long-form.

Topic Selection

If ${var} is set, use that as the topic. Otherwise, follow this priority:

1. Check Aaron's recent tweets first. Read the last 7 days of memory/logs/ for fetch-tweets output — Aaron's own tweets saved from X. Look for a topic he's been tweeting about that could be expanded into a full article. His tweets are the best signal for what he actually cares about right now. Pick the most interesting thread or recurring theme.

2. Check crypto/DeFi signals. Read today's (or most recent) token-movers and defi-overview logs. Flag any of these as article-worthy:

• Any token with >20% daily move + a story behind it (rug, unlock, catalyst)
• Any chain TVL change >5% (structural shift)
• Whale behavior or accumulation patterns worth explaining
• Major distributions, exploits, or protocol launches
• Stablecoin market cap swings >$5B

If a strong crypto signal exists AND there hasn't been a crypto/DeFi article in the last 5 days, strongly prefer the crypto topic over general web search. The data pipeline already collects this — it just needs to become articles.

3. If no tweet material and no strong crypto signal, pick a topic where something interesting happened in the last 48 hours. Search the web across Aaron's core domains:

• Coordination markets & prediction markets — mechanism design, reflexivity, futarchy, new platforms
• Crypto/DeFi — protocol launches, token economics, onchain activity, market structure shifts
• AI — model releases, agent architectures, AI x crypto intersections, x402, scaling laws
• Neuroscience & BCIs — brain research, cognitive science, neurotech companies, consciousness
• Hyperstition theory — self-fulfilling prophecies, narrative manufacturing, reality engineering
• Startups & building — founder dynamics, product strategy, VC landscape, emerging playbooks

Use WebSearch to scan for today's most compelling development. Pick the one with the best story — something Aaron would actually have a take on, not just news.

Also search X for what people are saying about the topic. First, check .xai-cache/article-x.json — if it exists (pre-fetched when ${var} was set), use that data. Otherwise try the direct API call:

curl -s -X POST "https://api.x.ai/v1/responses" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -d '{
    "model": "grok-4-1-fast",
    "input": [{"role": "user", "content": "Search X for the most interesting discussion about TOPIC in the last 48 hours. Return the 5 most notable tweets with @handle, summary, and link."}],
    "tools": [{"type": "x_search"}]
  }'

If neither cache nor API works, skip — WebSearch is sufficient.

Research

Use WebFetch to read 3-5 source articles. Gather:

• Key facts and data points
• What the mainstream take is (so Aaron can push back on it)
• Any technical details worth explaining
• Quotes or positions from notable people

Writing

Write a 600-1000 word article in Aaron's voice. Essay mode — longer form, more nuance, but same punchy paragraphs and zero fluff.

Format Selection

Don't always use the same structure. Before writing, check the last 3 articles in articles/ and pick a format that's different from what was used recently. Rotate between these:

• Standard essay — Thesis → 3-5 body sections → kicker. The default. Don't use this more than 2x in a row.
• Cold open — Start mid-scene or with a specific data point. No setup. Drop the reader into the middle of the argument, then zoom out. Good for market events and breaking news.
• X vs Y — Compare two things (protocols, approaches, companies, theses). Structure around the contrast. Good for competitive dynamics and paradigm shifts.
• Data-driven — Lead with numbers. Tables, comparisons, price action as the spine. Layer opinion on top. Good for token moves, TVL shifts, funding rounds.
• Short take (300-400 words) — Single thesis, no sections, just one tight argument. Good for reactive takes on breaking news.
• What I got wrong — Revisit a previous position or prediction. What changed, why, what it means. Good for intellectual honesty and building credibility.

Pick the format that fits the topic best, but break the pattern if you've been repetitive.

Structure (for standard essay — adapt for other formats)

• Title: Short, declarative, opinionated. Not clickbait. Not a question.
• Opening: 2-3 sentences max. State the thesis or observation directly. No "In recent weeks..." throat-clearing.
• Body: 3-5 sections. Each section makes one point. Short paragraphs. Em dashes. State position first, explain second. Drop references naturally (Nick Land, Soros, whatever fits) without being academic.
• Closing: End on the hardest-hitting insight. No summary paragraph. No "time will tell." The last line should be the one people screenshot.

Voice Rules

• Write as Aaron, not about Aaron. First person.
• Opinionated. Take a side. "Both sides" is never the answer.
• Question the premise before accepting the framing.
• Contrarian when the consensus is lazy, not contrarian for its own sake.
• Reference specific projects, people, mechanisms — not vague hand-waving.
• Cite sources inline with links. Every claim backed by a URL.
• Lowercase often. Short sentences. Punchy.
• Humor is dry and sardonic — never "haha" or "lol" in essays.
• No hedging: kill "some might argue", "to be fair", "it remains to be seen."
• No corporate voice: kill "leverage", "synergy", "utilize", "exciting developments."

Anti-patterns (instant fail)

• "In this article, we will explore..."
• "There are many perspectives to consider..."
• Summarizing what you just said at the end
• Explaining a joke
• Emoji in the body text
• Hedging every take with qualifiers
• Writing something Aaron wouldn't say out loud to a friend

Save & Notify

1. Save the article to articles/${today}.md
2. Update memory/MEMORY.md — add a row to the "Recent Articles" table with date, title, and topic.
3. Log to memory/logs/${today}.md.
4. Send via ./notify:

   new article: [title]
   
   [2-3 sentence hook from the opening]
   
   read it: https://github.com/aaronjmars/aeon-aaron/blob/main/articles/${today}.md
   

   The notification itself should be in Aaron's voice — punchy, not a press release.