The Big Sip
Image: Anthropic
Anthropic launched Skills on 16 October 2025, allowing users to package instructions, code, and reference documents into folders.
Claude loads these folders when relevant to tasks. The feature works across Claude AI, Claude Code, the API, and Agent SDK without requiring separate configuration for each use.
Anthropic projects that the feature will help triple revenue to $26 billion by 2026.
The company is monetizing enterprise customization by selling the ability to teach Claude company-specific procedures, brand guidelines, and compliance rules.
Will enterprises build Skills libraries, or will adoption lag like many low-code platforms?
Will OpenAI's AgentKit—launched 10 days earlier at DevDay on 6 October—capture the same enterprise budgets before Anthropic's Q4 sales cycle gains traction?
[Analysis] Anthropic official blog on Skills use cases — Rakuten's AI general manager Yusuke Kaji reports eight-fold productivity improvement in finance workflows, completing tasks that previously required a full day in just one hour by having Claude process multiple spreadsheets, catch anomalies, and generate reports using company procedures.
The timing is notable: OpenAI launched AgentKit at DevDay on 6 October, Oracle announced its AI Agent Marketplace on 14 October, and Anthropic dropped Skills two days later. All three companies are selling customizable AI that follows company-specific rules. Anthropic projects $19 billion in growth, believing that organized folders will compete with OpenAI's developer toolkit.
Here’s The Brew
General-purpose AI requires company-specific instruction manuals.
Within hours of the 16 October launch, Anthropic positioned Skills as "composable" and "portable"—instruction folders that move between their products.
Instead of enterprises repeatedly explaining brand guidelines and compliance rules in prompts, they package that knowledge into folders once. Claude loads them when tasks match.
Anthropic charges nothing extra for Skills (included in Pro/Team/Enterprise plans), but the projected $19 billion revenue jump by 2026 suggests they're banking on retention. Once enterprises build Skills libraries, switching to OpenAI or Google means rebuilding from scratch.
AI companies claimed models would "understand context" and "learn your preferences." Skills require enterprises to document their procedures in organized folders. Enterprises pay for the ability to do this documentation efficiently.
One folder system became Anthropic's biggest revenue bet. Companies building Skills libraries create switching costs while gaining productivity improvements.
Two Sides, One Mug
Image: Anthropic
Pro: Skills solve enterprise problems by eliminating repetitive prompt engineering. Non-technical employees can package domain expertise once and reuse it across teams. This explains Rakuten's eight-fold productivity improvement and why 80% of Anthropic's revenue comes from 300,000+ business customers who need consistent AI behavior.
Con: Skills are organized into folders with markdown files. The $19 billion revenue projection assumes enterprises will build and maintain Skills libraries. Many customization tools see declining usage after initial adoption.
Our read: Anthropic is betting that enterprises prefer easier customization over avoiding vendor lock-in. Companies building Skills libraries create switching costs while gaining productivity benefits. The company that makes customization easiest may win 2026 enterprise spending.
Receipt of the Day
Reuters exclusive on Anthropic revenue
Why it shifts the read: First public disclosure that Anthropic projects $20-26 billion annual revenue by 2026 (up from $7 billion now), with 300,000+ business customers representing 80% of revenue and Claude Code alone hitting nearly $1 billion run rate since launching earlier this year, revealing AI's enterprise business model is selling customisation tools rather than raw intelligence.
Spit Take
"What once took a day, we can now accomplish in an hour."
— Yusuke Kaji, Rakuten AI GM, on Skills
Coffee Break Links ×3
Anthropic engineering blog — Technical deep-dive on Skills architecture using "progressive disclosure" design pattern, with code examples and best practices for building custom skills. Worth the click if you want to understand why organized folders matter more than you'd think for AI performance.
SiliconANGLE coverage — Clean explainer on how Skills work across Claude AI, Claude Code, and API with examples of pre-built capabilities for Excel, PowerPoint, and PDF manipulation. Good overview if you skipped the marketing site.
The Register's take — a skeptical analysis noting that Skills present malware risks and that Anthropic warns users to "thoroughly audit" skills from untrusted sources, highlighting the security trade-offs enterprises face when letting AI execute bundled code.
Join your team of caffeinated skeptics. ☕
Opinionated world news that respects your time.
One bold take, the best counter, and the receipt(s) that prove it (all in six minutes).
Mugshot Poll 📊
Q: Which AI customisation feature will enterprises actually use?
Before we wrap up today…
Does crypto/Web3 intrigue you?
Is worrying about losing money or getting scammed stopping you from investing?
This course is for you!
Click to share (2 referrals get you free access to your very own crypto mastery email course).
Everything you need to unlock crypto in 5 days (for beginners).
You can read yesterday’s newsletter on the Nutcracker Man here.
For the love of coffee, see you on Monday!
Enjoy your weekend, keep it caffeinated.
How did we do?
Thanks for reading!
Be sure to get your daily curse and coffee fix by hitting that subscribe button.