Google Just Released Three Gemini Models at Once, Claude Can Now Analyze Your AI Usage Habits, and Notion Agents Went Mobile: Top 3 AI Updates, Week of July 24, 2026

The pace of AI updates did not slow down this week.

In the fourth week of July 2026, three companies released major announcements almost simultaneously. Google unveiled three new Gemini models at once. Anthropic gave users a tool to look inside their own Claude usage patterns. Notion pushed its agents onto smartphones and handed them calendar control.

The common thread across all three: AI tools are moving beyond "features" and embedding themselves into habits and environments.

Table of Contents

1. Gemini 3.6 Flash: Not One but Three New Models
2. 3.5 Flash Cyber: A Cybersecurity Specialist Model
3. Claude Reflect: A Mirror for Your AI Usage Habits
4. Claude Microsoft 365 Write Tools: Email, Calendar, and OneDrive
5. Notion Agents iOS: Agents in Your Pocket
6. Notion Calendar Tools: Let Agents Manage Your Schedule
7. What These Three Updates Signal

1. Gemini 3.6 Flash: Not One but Three New Models

On July 21, 2026, Google DeepMind announced three Gemini models simultaneously: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber.

The intent is clear: each model addresses a distinct type of demand — general performance, lightweight efficiency, and security specialization.

Gemini 3.6 Flash: Google's New Workhorse

Gemini 3.6 Flash is what Google calls its "workhorse model" — the general-purpose centerpiece of the new lineup. Compared with the previous 3.5 Flash, three things changed:

• Up to 17% reduction in token usage: The same tasks get done with fewer tokens, which translates directly into lower API costs for developers and teams.
• Improved coding and knowledge-work performance: Code generation, document analysis, and complex reasoning tasks all show improvement.
• Stronger multimodal handling: Stability in tasks that combine text, images, and audio has improved.

3.5 Flash-Lite: Maximum Efficiency at Minimum Cost

3.5 Flash-Lite is the most cost-effective model in this generation. It's designed for high-volume processing, fast-response applications, or any service where cost is the primary constraint. For educational platforms or solo developers building products where both speed and price matter, this is the natural choice.

2. 3.5 Flash Cyber: A Cybersecurity Specialist Model

3.5 Flash Cyber occupies the most distinctive position among the three: it is a specialized model fine-tuned specifically for detecting and fixing cybersecurity vulnerabilities.

If a general-purpose model is a generalist who covers every domain adequately, 3.5 Flash Cyber is a specialist trained intensively on security domain data. It handles vulnerability classification, patch suggestions, and code auditing at a competitive price point.

"Assigning security tasks to a general model makes it hard to guarantee consistent results. A model fine-tuned on security domain data understands context more accurately."

What These Models Mean for Educators and Content Creators

The launch of three models is not just a developer story.

• Google AI Studio users can access 3.6 Flash immediately.
• Gemini app subscribers (Pro and Ultra) can select 3.6 Flash from the model options menu.
• Education platform developers: delivering the same performance at 17% lower cost meaningfully improves service sustainability.
• Independent researchers and solo creators currently using the API: switching to 3.5 Flash-Lite is a viable cost-reduction strategy.

3. Claude Reflect: A Mirror for Your AI Usage Habits

Anthropic added a built-in analytics dashboard called "Reflect" to Claude. It visualizes the topics you've discussed, your usage patterns, and the types of tasks you use AI for.

Reflect is not just a usage counter. It answers questions like: What am I primarily using Claude for? When do I use it most? Which kinds of tasks do I lean on AI for, and which do I handle myself?

What Reflect Shows You

• Topic distribution: A breakdown of what categories your conversations fall into — writing support, coding help, research, creative work, and so on.
• Time-of-day usage patterns: Whether you lean heavily on Claude in the mornings or evenings, and how weekdays compare to weekends.
• Task type classification: Whether your usage skews toward quick question-and-answer, repeated task automation, or deep exploratory conversation.

Why This Matters

Using AI tools well starts with knowing how you actually use them. There is a meaningful difference between a vague sense of "I use Claude every day" and knowing "I rely on Claude primarily for morning writing sessions, and I almost never use it for coding." Reflect makes the second kind of knowledge accessible.

It can help you spot blind spots in your usage, identify areas of high AI dependency versus low dependency, and make intentional choices about how you want to shift that balance.

"Changing a habit requires seeing the habit first. Reflect is the first tool that makes AI usage habits visible."

An Educator's Perspective

For teachers and educational administrators, a tool like Reflect carries particular significance. As AI literacy education grows, teaching students to monitor and reflect on their own AI usage is becoming a core competency. Reflect models, at the individual user level, exactly what that kind of self-monitoring looks like in practice.

4. Claude Microsoft 365 Write Tools: Email, Calendar, and OneDrive

Alongside Reflect, Anthropic added write tools to Claude's Microsoft 365 connector. Claude can now draft and send email, manage calendar events, update mailbox settings, and create or edit files in OneDrive and SharePoint.

If the previous Microsoft 365 integration was primarily read-only, this update turns Claude into an active workflow executor.

The scenario of asking an AI assistant to "check this week's meeting schedule and decline the ones I can't make" is now available inside the Microsoft ecosystem.

5. Notion Agents iOS: Agents in Your Pocket

Notion launched its Agents feature as an iOS app. You can now run Notion agents, check their results, and issue follow-up instructions from a mobile device.

Until now, Notion Agents worked properly only in a desktop web browser. The iOS app extends the context in which agents can be used — from the desk to wherever you happen to be.

What You Can Do on iOS

• Monitor the status of ongoing agent tasks
• Issue new instructions to agents
• Review and approve outputs generated by agents
• Check auto-generated summaries of meeting notes

Why mobile matters here is about context. On a commute, between appointments, or during a short break, you can check what your agents have done and tell them what to do next — without waiting to get back to a desk.

Other Changes in This Notion Update

• Audio upload: You can now upload audio files directly to AI Meeting Notes. The recording gets transcribed and summarized automatically — just like a live meeting.
• Improved speaker identification: The system now identifies speakers based on which microphone is active, so follow-up actions are assigned to the right person.
• Agent duplication: Any custom agent can be duplicated in one click and adjusted for a different team or use case without rebuilding from scratch.
• New AI model options: Agents now support Opus 4.8, Grok 4.3, and the open-weight GLM 5.2.

6. Notion Calendar Tools: Let Agents Manage Your Schedule

Notion gave its agents direct calendar authority in this update. Agents can now manage schedules, join calls, send invites, and find open meeting times — without you doing any of the coordination yourself.

The distinction from previous calendar integrations is directional. Before: read. Now: write and act.

Real Scenarios

1. Meeting coordination: "Schedule a team meeting for next Tuesday afternoon and send calendar invites to everyone."
2. Conflict resolution: "Check this week's schedule for overlaps and draft a cancellation for the lower-priority one."
3. Recurring meeting setup: "Create a monthly review on the first Monday of each month going forward."

"The most time-consuming part of scheduling is the back-and-forth of finding a time that works. When an agent reads the calendar and handles the coordination, that entire process disappears."

7. What These Three Updates Signal

Placed side by side, this week's updates reveal a single direction.

First, AI is shifting from "tool" to "environment." Gemini 3.6 Flash being cheaper and faster means more than incremental improvement. When AI gets less expensive to run, the threshold for embedding it into services drops. AI stops being an exceptional feature and starts becoming the default.

Second, a layer of self-awareness about AI usage has appeared. Claude Reflect is not just a statistics screen. It is a meta-layer that lets users see how they use AI — how much, for what, toward what ends. In AI literacy education, knowing your own usage patterns is the starting point of using AI well. The educational implications here are real.

Third, AI agents are moving. Notion Agents iOS shows that the execution environment for agents is breaking free of a fixed desktop. Instead of "setting up an agent and waiting," the pattern becomes "checking in and giving instructions while on the move."

Where these three directions converge — cheap AI infrastructure, self-aware users, agents you can direct from anywhere — is the concrete shape of AI productivity in the second half of 2026.

Closing

Google split its models three ways this week. Differentiating into general, lightweight, and security-specialized tracks reflects the reality that AI demand is not uniform. Anthropic held up a mirror: making it possible to see how you use AI is the precondition for using it better. Notion moved its agents to your phone. An agent now authorized to handle your calendar can get things done outside the meeting room.

All three changes point toward AI occupying more space in daily life. How we receive that is still our question to answer.

Further Reading

• Claude Code Now Defaults to Sonnet 5, Gemini Deep Research Reads Your Files
• NotebookLM Became Gemini Notebook, Claude Artifacts Now Run Live

Which of these three updates would you try first? Leave a comment!

Sources

• Google releases three new Gemini models — but no 3.5 Pro — TechCrunch
• Gemini Updates by Google — July 2026 — Releasebot
• Anthropic's new Claude feature is quietly selling you on AI — TechCrunch
• Claude Updates by Anthropic — July 2026 — Releasebot
• Notion Release Notes: July 1, 2026 — Notion 3.6
• Notion Release Notes — July 2026 Latest Updates — Releasebot