The AI Instruction Insider June 2025

Claude Sonnet 4: The new benchmark for clarity and context

What is the AI news?

Anthropic just released Claude Sonnet 4, and it’s already making waves for its exceptional performance in generating clear, coherent content. It’s the latest mid-tier model from Anthropic’s Claude 3.5 family, but don’t let “mid-tier” fool you - it’s setting a new standard. Many are calling it the best model currently available for writing tasks, thanks to its standout abilities in context comprehension and lucid writing.

Why it matters

Context is everything in technical documentation. Whether you’re writing API docs, user guides, or internal SOPs, losing track of context can turn your doc into a confusing mess. Claude Sonnet 4 shines here - it retains context better than its peers, making fewer assumptions and maintaining clarity throughout long or complex documents. For technical writers juggling versioning, product updates, and multiple audiences, this model is like having a super-sharp editor that never tires.

How it helps you

Claude Sonnet 4 is especially useful when you're refining dense or fragmented material. Need to summarise a complex process? It keeps the nuance intact. Updating legacy documentation? It understands what to keep, what to clarify, and what to cut. Even better, it generates output that often requires less post-editing than what you’d get from GPT-4 or Gemini.

You can use Claude Sonnet 4 to:

• Reorganise the jumbled content logically.
• Write first drafts that sound like final drafts.
• Interpret and rewrite code comments or developer notes into user-friendly explanations.
• Maintain a consistent tone and terminology across long documents.

If you haven’t yet, give Claude Sonnet 4 a test run in your workflow. You might find it becomes your go-to assistant for everything from drafting to revising. Whether you're knee-deep in software docs or building a new help centre from scratch, this model makes technical writing smoother and smarter.

Guidde vs. Scribe: What’s the right tool for AI-powered instructions?

In the age of AI-driven documentation, tools like Guidde and Scribe offer new ways to capture and share procedures. But which tool fits which use case? At INSTRKTIV, we tested both to better understand their capabilities and their limitations.

What is Guidde?

Guidde is a visual-first tool that lets you record your screen and instantly generate AI-powered how-to videos. You click through a workflow, and Guidde breaks it down into steps, automatically generating visuals and written instructions. The result is a polished video guide you can share with your team, customers, or audience.

Key features include:

• AI-generated voiceovers in multiple languages
• Click-to-capture step-by-step instructions
• Export to video, slide deck, GIF, or embed in help centers
• Integrations with tools like Notion, Zendesk, and Salesforce
  
  

It’s particularly strong for customer support teams or SaaS companies looking to create engaging onboarding content quickly and at scale.

What is Scribe?

Scribe works similarly but focuses more on text-based documentation. When you record a process, it automatically captures clicks and inputs, then generates a step-by-step guide with screenshots and text. It’s ideal for internal documentation, SOPs, or onboarding manuals.

Output formats include:

• PDF, Word, and HTML
• Confluence integration
• Embedded scroll guides or slide-like formats (Pro plan)

Our experiment with Scribe

At INSTRKTIV, our technical writer Erwin tested Scribe by documenting an internal process. He recorded a basic procedure and edited the steps by adding extra descriptions and instructions. The platform made this straightforward.

The output could be exported in multiple formats, and we appreciated how the content updated automatically when changes were made - especially useful for evolving internal processes.

INSTRKTIV’s take

Both tools are powerful, but they serve different purposes:

• Scribe shines in creating fast, clear, and structured text-based documentation, especially for teams managing sensitive or complex internal knowledge. Its integration with tools like Confluence and strong privacy features make it a great fit for large teams or regulated industries.
• Guidde, on the other hand, excels when the goal is to create engaging video instructions, such as for product onboarding or external help content. Its AI voiceover and visual capabilities go beyond what Scribe offers in that domain.

However, neither tool is (yet) suitable for creating full user manuals - the kind of complex, legally compliant documentation that we typically produce at INSTRKTIV. Such manuals require contextual information, structured frameworks, risk communication, and localisation support that go far beyond what either platform currently offers.

Bottom Line

• Choose Guidde for fast, engaging visual guides with AI voiceovers
• Choose Scribe for internal, structured text-based documentation
• For complete user manuals: stick with purpose-built tools (like our own AI Agent) combined with technical writing expertise

Want to know more about our experiments or see how our own AI tools compare? Just reach out, we’re always happy to share what we’ve learned.

More June highlights: AI is building, browsing, and (almost) brewing coffee

What is the other AI news?

While you were having your morning coffee, AI may have already finished coding an entire web app. June brought a wave of innovation in artificial intelligence, with Google leading the charge during its I/O 2025 event. Several major tools were announced:

• Veo 3 – expanding the capabilities of AI video generation
• Gemini 2.5 Pro – delivering more power, better reasoning, and broader context handling
• Flow – enabling AI integration directly into enterprise workflows

These tools represent a meaningful step forward in making AI not only smarter but also more useful in practical, day-to-day business environments.

AI’s role in software development has also taken centre stage. Rather than assisting developers, AI is now generating full-stack applications, handling backend logic, and automating code review. Tools such as GitHub Copilot, Gemini Code Assist, and Claude’s workspace features are accelerating how developers work.

Meanwhile, in content creation, platforms like Simplified and Conteflow are combining text, video, and image generation into unified workflows. Although they are not yet optimised for technical documentation, they demonstrate how AI is moving toward supporting high-volume content production across functions.

Perhaps most notably, AI agents are now capable of performing actual tasks. The Open Computer Agent (OCA) from Hugging Face enables AI to browse, click, and type within virtual environments. Mistral’s newly launched platform lets developers build task-performing agents that execute real-world actions.

Why it matters

These developments mark a shift in how AI contributes to work:

• AI is no longer just assisting with development - it’s becoming part of the core toolkit across teams and systems.
• Creating content is now quicker and easier, with tools that help teams scale without losing quality.
• AI agents are starting to do more than just reply - hey’re actually taking action and carrying out real tasks.
• For industries such as compliance, documentation, and support, this presents new opportunities for automation and consistency.

How it helps you

You can apply these developments in multiple ways:

• Speed up development – Use AI tools like Copilot or Gemini Code Assist to generate production-ready code and automate testing.
• Streamline content creation – Platforms like Conteflow allow your marketing or training teams to produce consistent assets quickly.
• Automate internal tasks – Use task-performing agents for document reviews, data extraction, or compliance workflows.

Whether you’re a developer, writer, or operations lead, now is a good time to explore how these tools can reduce manual effort and increase efficiency across your team.