Read 1M Tokens Without Losing the Thread, Copy a Notebook in One Click, Deploy Code Without a Server: 3 AI Picks for August 2026

Three updates landed at the end of August — and each one quietly lowers a barrier that used to require specialist skill.

One is an AI model that lets you decide, per request, how hard it should think. Another is a feature that distributes dozens of hours of curated research to an entire team with a single copy. The last is a developer platform that lets you run automation code inside Notion without ever touching a server. The common thread: something that only experts could do before just got easier for everyone.

Claude Opus 5: 1M Tokens Without Dropping a Thread, and You Set the Effort
What the Effort Toggle Actually Changes: Trading Cost for Intelligence
Gemini Notebook: Copy a Notebook, Auto-Update Its Sources
Notion Workers: Deploy Code on Notion's Infrastructure, No Server Required
What These Three Updates Point To

1. Claude Opus 5: 1M Tokens Without Dropping a Thread, and You Set the Effort

Anthropic released Claude Opus 5 on July 24, 2026. In August, text watermarking was added for EU AI Act compliance. It's the largest leap the Opus line has seen — and the numbers explain why.

Opus 5's context window is 1 million tokens. That isn't just a big number. The model scores 100% on the Needle in a Haystack benchmark — it can retrieve a fact buried anywhere in a 1M-token document without ever losing it. One book is roughly 100,000 tokens. Opus 5 reads ten books simultaneously and knows exactly where everything is.

The Effort Toggle: Five Levels of Thinking

Opus 5's most distinctive feature is the Effort Toggle. For every request, you specify how deeply the model should reason before answering.

Level	Best For	Relative Cost
Low	Quick summaries, fast answers	Lowest
Medium	General tasks	Moderate
High	Complex analysis, coding	High
XHigh	Deep reasoning tasks	Very high
Max	Tasks that demand peak capability	Highest

Thinking tokens bill as output tokens, so lower effort levels genuinely reduce cost. Using Max for a quick spell-check is like booking a Michelin-starred chef to make toast — possible, but wasteful.

Pricing and Performance

Input: $5 per million tokens (same as Opus 4.8)
Output: $25 per million tokens (same as Opus 4.8)
Performance: New state-of-the-art on Frontier-Bench and GDPval-AA
Available on: Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry

Same price as the predecessor, near Claude Fable 5–level intelligence. Anthropic's own framing: "Fable 5 intelligence at half the price."

Text Watermarking (August Update)

In August, Claude output gained cryptographic text watermarking to comply with the EU AI Act — the first mechanism that lets you technically prove a piece of text was generated by Claude.

"The first time I used Opus 5, there was no awkward warmup period. I dropped an entire contract into the context and asked where a specific clause appeared and what surrounded it. It never missed once."

2. What the Effort Toggle Actually Changes: Trading Cost for Intelligence

The Effort Toggle isn't just a settings knob. It's a new interface for how you use AI.

Until now, AI model pricing was flat: every query, same model, same cost. With Opus 5, that changes. You can calibrate how much compute the model spends based on how complex the task actually is.

Practical Patterns for Educators and Solo Builders

Fast draft review: Low — scan for typos and flow issues, nothing deeper
Research synthesis: Max — fully digest a stack of papers and distill the core arguments
Code review: High — find potential bugs and security issues without skipping anything
Brainstorming: Medium — quick pivots across many angles, cost kept reasonable

For developers using the API: build a pipeline that classifies incoming queries by complexity and routes each one to the right effort level automatically. A simple FAQ hits Low; a legal document review hits High. The routing pays for itself immediately.

3. Gemini Notebook: Copy a Notebook, Auto-Update Its Sources

In August, Google added two significant features to Gemini Notebook (formerly NotebookLM): full notebook copying and automated source updates through Workspace Studio.

Gemini Notebook's core idea is unchanged: gather sources — documents, links, YouTube videos, PDFs — into one notebook, and let the AI answer questions using only those sources. Unlike a general AI that searches the whole web, it stays within the materials you trust. That constraint is its strength.

Full Notebook Copying (August 17)

You can now copy an entire notebook in one action. What gets copied:

All sources: links, Drive files, YouTube URLs, web pages
Studio artifacts: Audio Overviews, Video Overviews, Study Guides, Flashcards, Quizzes, Slide Decks
What's excluded: personal chat history, user notes (privacy protected)

Why does this matter? Until now, Gemini Notebook was a personal research tool. Building a notebook with dozens of sources and a full suite of study materials takes hours. Now that work can be distributed to a whole class or team in one click. A course-unit notebook shared with students; a project research base handed to a new team member the moment they join.

Automated Source Updates (August 7)

The Workspace Studio automation flow now has an "Add a source to Gemini Notebook" step. You can automatically push text, Drive file links, YouTube URLs, or web URLs into a notebook on a recurring schedule.

Set a weekly newsletter to automatically land in your notebook as a new source. The manual chore of keeping a notebook current disappears.

"The most tedious part of building a research notebook was keeping the sources up to date. Once that's automated, the notebook becomes a living database instead of a snapshot."

For Educators

Create per-unit notebook templates and distribute them to an entire class
Set up weekly news auto-collection to automate discussion prep
Share student research notebooks with teachers for feedback — just copy, no rebuild

4. Notion Workers: Deploy Code on Notion's Infrastructure, No Server Required

In August, Notion opened Workers as a public beta — a platform for running custom code directly on Notion's infrastructure. Credit-based billing started August 11; the developer portal landed in the sidebar on August 19.

The one-line description of Notion Workers: serverless functions for Notion. The concept is similar to AWS Lambda or Vercel Functions, but the runtime lives inside your Notion workspace.

What Workers Can Do

Database sync: Pull data from external APIs and automatically update Notion databases
Custom agent tools: Define functions that Notion's AI agents can call
Webhook triggers: React to external events (GitHub commits, Slack messages) and execute Notion actions
Automation extensions: Implement logic too complex for Notion's built-in automations

How to Deploy

# Deploy a Worker with the Notion CLI
npm install -g @notion-hq/cli
notion worker deploy my-worker

Write the code, run the deploy command, done. No server configuration, no environment variable management, no scaling concerns. Notion runs it in a sandboxed environment.

Pricing

Beta period: Free through August
From August 11: Credit-based billing at $10 per 1,000 credits
Access: Business and Enterprise plans for deploy and management; CLI available on all plans

Developer Portal (August 19)

A Developer section now lives in the Notion sidebar. Manage deployed Workers, execution logs, connected integrations, and personal access tokens in one place.

"You write the code and Notion runs it. Not having to think about a server is the whole point. This is the moment Notion stops being a note app and becomes a development platform."

Practical Uses for Solo Builders and Educators

Student enrollment sync: Check an external payment system and auto-update a Notion database with enrollment status
Content publishing pipeline: Auto-commit Notion drafts to GitHub on a schedule
Submission organizer: Receive form responses and auto-classify them into a tagged Notion database

5. What These Three Updates Point To

Set the three updates side by side and one trend becomes clear.

First, users can now control how much resource AI consumes, at the request level. Opus 5's Effort Toggle is the first time you can decide "how deeply should the AI think about this" on a per-request basis. That's not a feature — it's a change in the fundamental contract between user and model.

Second, research assets are beginning to flow from individuals to teams. Gemini Notebook's copy feature turns dozens of hours of curated research into an instantly shareable asset. The structure that locked knowledge to individuals is starting to crack.

Third, the line between developer and non-developer is blurring again. Notion Workers lets people who don't know what a server is run automation code inside Notion. Complex data pipelines without infrastructure knowledge is now a real scenario, not a future one.

The people at the intersection of all three — those who choose the right AI depth for each task, share research assets with their teams, and extend Notion with code — will be the real beneficiaries of AI productivity in the second half of 2026.

Wrapping Up

Claude Opus 5 let you control how deeply it thinks. Gemini Notebook let you distribute what you've built. Notion Workers let you run code without a server.

All three hand something that used to require expertise to people who don't have it yet. The question is no longer "how do I use these tools" — it's "what will I build with them?"

Related Posts

Which of the three updates would you apply to your workflow first? Let me know in the comments!

Sources

Table of Contents