🪨

Feed your second brain
from your phone.

You read on your phone. Your knowledge base lives in Obsidian. Petrify bridges the gap — one tap turns any link into a structured atomic note in your vault.

Get Petrify — $5 One-time purchase. No subscription. Yours forever.
📱 Read on your phone
→
🔗 Share the link
→
⚙ AI reads & structures
→
🪨 Note in your vault

Populate your vault, not your bookmarks

Stop saving links you'll never revisit. Petrify reads them for you and deposits structured knowledge.

⚡

One-tap capture

Share any URL from Safari, Chrome, X, or any app. Petrify does the rest.

📌

Claim-titles

Every title is a substantive assertion you can build on — not a generic label like "AI News".

🪨

Structured layers

YAML frontmatter, tags, summary, key ideas, quotes, and connection suggestions. Every time.

🐦

Twitter/X native

Extracts tweets, threads, and linked articles — even X's long-form posts.

🌐

Any language

English by default. One config field to get notes in Spanish, French, German, or any language.

🔑

Your key, your data

Bring your own Anthropic API key. No middleman, no markup. ~$0.01 per note.

From link to knowledge

Share an article while scrolling on your phone. 15 seconds later, this lands in your vault.

---
source: https://example.com/ai-progress
date: 2026-04-11
tags: [ai, scaling, prediction]
status: raw
author: external
---

# Reasoning models have collapsed the competitive advantage of proprietary data

The latest generation of reasoning models demonstrates that massive proprietary datasets no longer guarantee a competitive edge. Open-weight alternatives now match or exceed closed models across most standard benchmarks.

This shift suggests the moat in AI has moved from data accumulation to inference-time compute and architectural innovation.

## Key Ideas
- Open-weight models match proprietary ones on most benchmarks
- The moat has shifted from data to inference-time compute
- Fine-tuning on domain data yields diminishing returns
- Smaller labs can now compete without billion-dollar datasets

## Notable Quotes
> "The data advantage era lasted exactly four years."

## Possible Connections
- Scaling laws and compute efficiency debates
- Open-source vs proprietary AI ecosystem dynamics

Simple pricing, permanent notes

Pay once. Build your knowledge base forever.

$5
One-time purchase · yours forever
  • iOS Shortcut file (1-tap import)
  • Setup guide (English & Spanish)
  • Works on iPhone, iPad & Mac
  • Configurable language & tags
  • Free updates
Buy Petrify
BYOK (Bring Your Own Key): You use your own Anthropic API key and pay them directly for AI processing. Typical cost: $0.01-0.03 per note. Most users spend under $2/month. No subscriptions, no middleman.

FAQ

What do I need?

iPhone with iOS 16+, Obsidian with iCloud Sync, and an Anthropic API key (free to create at console.anthropic.com, pay-per-use).

How much does each note cost?

About $0.01-0.03 via Claude's API. You pay Anthropic directly. A heavy user (10 notes/day) spends ~$5-10/month. Most people spend less than $2.

Does it work with Obsidian Sync (paid)?

Petrify saves to iCloud Drive. If your vault uses Obsidian's paid Sync, you'll need to move files manually or switch to iCloud Sync.

Can I use it on iPad or Mac?

Yes. The shortcut works on any Apple device with Shortcuts. On Mac it appears in the Share menu.

Is my API key safe?

Your key stays on your device and is sent directly to Anthropic. We never see or store it.

What AI model does it use?

Claude Sonnet 4.6 — the best balance of quality and cost for structured note generation.