Adam here!

OpenAI is building its own chip to support the cost and scale of AI. Samsung is rolling out ChatGPT and Codex across its workforce. Codex can now learn workflows by watching them once. Claude and Gemini are moving into Slack, screens, and software interfaces. And Figma is bringing design and code closer together.

AI does not just help you write faster. It helps you listen faster.

This week inside AI Systems Lab, we looked at how to use AI to turn customer language into better marketing.

The main idea was simple: your customers have probably already written your best ads. Their reviews, sales calls, support tickets, FAQs, objections, and offboarding conversations are full of the exact words they use to describe their problems, goals, doubts, and buying triggers.

And that idea connects well with the bigger AI stories this week. AI is becoming less of a separate tool you open and more of a working layer across the systems businesses already use.

ANTHROPIC

Claude Fable 5 and Mythos 5 Are Coming Back After Export Controls Were Lifted

Anthropic says it has received notice that the US Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. The company said it will begin restoring access and share another update soon. Reuters also reported that the US Commerce Department lifted the controls less than three weeks after the original suspension.

Earlier in June, Anthropic disabled access to Fable 5 and Mythos 5 after receiving a US government export control directive requiring the company to suspend access to the models by foreign nationals. Anthropic said at the time that it had to disable the models for all customers to ensure compliance.

Now, access is expected to return after the restriction was lifted.

OPENAI

OpenAI’s First Chip Shows the AI Race Is Also an Infrastructure Race

OpenAI and Broadcom introduced Jalapeño, OpenAI’s first custom AI inference chip.

Jalapeño is designed for inference, which is the process of running AI models after they have already been trained. Every time someone uses ChatGPT, Codex, an API call, or an AI agent, inference is part of what happens behind the scenes.

Most businesses will not need to think directly about AI chips. But the impact may show up in the tools they use.

Better infrastructure could eventually mean:

  • Faster AI responses

  • Lower operating costs

  • More reliable AI tools

  • Better access to advanced models

  • More specialised AI products for businesses

The practical takeaway is that AI is becoming a full-stack industry. The tools businesses use on the surface are increasingly shaped by infrastructure decisions happening behind the scenes.

SAMSUNG

Samsung Is Rolling Out ChatGPT and Codex Across Its Workforce

Samsung Electronics is rolling out ChatGPT Enterprise and Codex to employees in South Korea, as well as employees across its Device eXperience division worldwide.

The rollout will support teams across areas like R&D, software development, product development, marketing, manufacturing, and corporate functions.

This is not just a small AI pilot for one team. It is a large enterprise rollout across both technical and non-technical roles.

This is what AI adoption looks like after the pilot phase.

Many businesses have already had people experimenting with ChatGPT or other AI tools individually. Samsung’s rollout shows the next stage: making AI available across departments and building it into the way teams work.

Codex is also no longer just a coding tool for developers. It can help non-technical teams turn ideas into internal tools, websites, automated workflows, and working software.

OPENAI

OpenAI’s Record & Replay Lets Codex Learn a Workflow by Watching It Once

OpenAI introduced Record & Replay for Codex, a feature that lets users record themselves doing a workflow on their Mac and turn it into a reusable skill.

Instead of writing a long prompt explaining every step, users can show Codex how the task is done. Codex then turns that demonstration into an editable skill that can be reused later.

This could be useful for repetitive admin, operations, content, and reporting workflows.

Pick a workflow that you already know how to complete. Record & Replay works best when the steps are stable and the success criteria are clear.

  1. Open Plugins in the Codex app.

  2. Open the + menu.

  3. Select Record a skill.

  4. Review the suggested prompt, give Codex any helpful context, and submit it.

  5. When Codex asks for permission to record your actions, approve the request once you are ready to demonstrate the workflow.

  6. Perform the workflow on your Mac.

  7. When you are done, stop recording from the menu bar, overlay, or tell Codex that you are done.

If a process is easier to show than explain, it may become a good candidate for AI-assisted workflow automation.

ANTHROPIC + GOOGLE

AI Agents Are Moving Into Slack and Screens

Anthropic launched Claude Tag, a Slack-based workflow where teams can mention Claude inside a channel and assign tasks directly from the conversation.

Google also added computer use capabilities to Gemini 3.5 Flash, allowing developers to build agents that can interact with browser, mobile, and desktop interfaces using screenshots and actions like clicking, typing, and scrolling.

For teams, this is a good time to look at where an AI assistant could sit inside existing workflows.

The important part is setting boundaries. Before giving an AI agent access to tools, teams should be clear about what it can do, what it should only draft, and what still needs human approval.

FIGMA

Figma Is Bringing Design and Code Closer Together

Figma announced a major update at Config 2026, including code layers, Figma Motion, shaders, generative plugins, and new tools directly on the canvas.

One of the bigger updates is code layers, which bring code into the same collaborative space where design work happens.

Teams can use this kind of update to speed up early-stage product, website, and campaign work.

For example:

  • Turn design concepts into more interactive prototypes

  • Bring developers into the design process earlier

  • Test landing page or product ideas faster

  • Use AI to create visual effects or interface variations

  • Reduce the back-and-forth between design and build teams

For agencies and marketing teams, this could make campaign prototyping and website iteration faster, especially when design and development need to move together.

TWEETS OF THE WEEK
MORE AI IN ACTION
Instagram post
Instagram post
Instagram post
Instagram post
Instagram post
Instagram post
DISCOVER THESE AI TOOLS AND APPS
  1. AgentDock - An open-source platform for creating and deploying AI agents and automated workflows, with a visual builder, integrations, and modular setup.

  2. Caddie - An AI copilot for sales calls that can help with live research, recall, and answers while the conversation is happening.

  3. AtomWords - A free visual dictionary for AI image prompts, helping users find better words for styles, lighting, colours, moods, camera angles, and compositions.

  4. Aside - An AI browser built to work across logged-in websites and help complete more complex browser-based tasks.

  5. Plotly Cloud - A tool that helps turn datasets into interactive data apps, dashboards, and reports in minutes.

  6. Vokal - A collaboration space where teammates and AI agents can work together using shared channels, tasks, docs, tools, memory, and a knowledge base.

  7. Exa Connect - A tool that connects AI agents to public and private data sources, helping them search and retrieve better information for workflows.

  8. Cleanaudio - An AI tool that removes background noise from audio and video recordings.

  9. Audjust AI - A web-based audio editor and music generator that can resize tracks, create loops, convert audio to MIDI, and generate songs.

  10. CanIRun.ai - A hardware checker that shows which AI models your computer can run based on your GPU, CPU, RAM, and WebGPU support.

Keep Reading