Sharp in Five
Archive

Friday, October 2, 2026

Share this briefWhatsAppLinkedInX

The Brief

Meta integrated its Muse AI model into new smart glasses, enabling the assistant to process live video of your surroundings. Mobile professionals can now retrieve information and interact with physical spaces hands-free instead of pulling out a screen.

  • Google launches Gemini 4 Argon. The company released a new model built to match top-tier systems, bringing more reliable reasoning to complex tasks inside Google's software products. Source
  • Unrestricted cyber capabilities. Anthropic research shows the new GLM-5.3 model can autonomously build complex cyber exploits, requiring IT managers to prepare for highly automated phishing threats. Source
  • Claude Code task costs. Since our earlier report on Opus 5.5, Anthropic broke down the exact compute costs for automated engineering tasks so managers can accurately budget. Source

Try this at work — for operations and finance

  1. Open ChatGPT and upload a blank regulatory form or compliance application your team fills out regularly.
  2. Paste your standard company background data into the chat and instruct the AI to draft responses for every required text field.
  3. Review the completed draft for missing context and verify all generated dates and financial figures against your internal records before submission.
The details

Hands-on

2 picks
  1. Claude Code

    How to build and use custom mods in Claude Code

    Anthropic has released a guide on creating 'mods,' which are custom hooks that extend the capabilities of the Claude Code terminal tool. Technical teams can use these to automate specific parts of their programming environment or integrate specialized workflows directly into the AI coding assistant.

    Claude Developer BlogRead
  2. Claude

    When to choose Claude Sonnet 5.5 over Opus for development

    This guide compares the performance and costs of the Sonnet 5.5 model against the high-end Opus version. It helps developers decide which model provides the best balance of intelligence and price for specific coding or automation tasks.

    Claude Developer BlogRead

The AI race

5 items

  1. Meta announces Muse AI model for smart glasses and new hardware

    At the Connect 2026 event, Meta introduced its Muse AI model integrated into wearable glasses alongside new VR hardware. This signals a shift toward ambient AI assistants that can see what you see, potentially changing how mobile professionals interact with digital information hands-free.

    Meta NewsroomRead the full article

  2. Ideogram 4.5 introduces precise image editing and 2K resolution

    Ideogram 4.5 allows users to modify specific parts of an image while keeping the rest of the composition perfectly intact. This makes the tool much more viable for professional design tasks where iterative feedback and minor adjustments are required without starting from scratch.

    The DecoderRead the full article

  3. GLM-5.3 raises security concerns over autonomous cyber exploit capabilities

    Anthropic research highlights that the new GLM-5.3 model can autonomously build complex cyber exploits without the safety guardrails found in other frontier models. Knowledge workers and IT managers should be aware of the increased potential for automated phishing and security threats as these unaligned models become public.

    Newsletter radarRead the full article

  4. Google launches Gemini 4 Argon to compete at the AI frontier

    Google has introduced Gemini 4 Argon, a new model designed to match the capabilities of the world's most powerful AI systems. For knowledge workers, this promises more reliable reasoning and better handling of complex professional tasks within Google's ecosystem.

    Newsletter radarRead the full article

  5. Amazon releases Strands Decider 2B for automated decision making

    Amazon Web Services has released Strands Decider 2B, a specialized model designed for logical decision-making tasks. While smaller in size, it aims to provide efficient reasoning for automated business workflows and logic-heavy applications.

    TechCrunch AIRead the full article

AI at work

5 items

  1. Social club cuts grant application time by 90% using ChatGPT

    The Den, a social club group, is using ChatGPT to reduce the time spent on grant applications from three days to just two hours. By automating complex paperwork like liquor licenses, the team saves 10–15 hours per week. This case study demonstrates how knowledge workers can reclaim significant time by using AI for repetitive regulatory and administrative drafting.

    OpenAI NewsRead the full article

  2. Shopify Canvas lets users build online stores via conversational chat

    Shopify’s new Canvas tool integrates their Sidekick AI agent to allow merchants to build and edit entire websites just by describing what they want. You can now see layout changes happen in real time as you chat, removing the need for manual design or coding skills.

    TechCrunch AIRead the full article

  3. Albertsons uses ChatGPT Enterprise to automate retail tasks and shopping

    Grocery giant Albertsons is deploying ChatGPT Enterprise and OpenAI's API to help internal teams work faster and improve customer experiences. This case study demonstrates how large organizations are moving past experimentation to integrate AI directly into supply chain and retail operations.

    OpenAI NewsRead the full article

  4. OpenAI partners with SBDC to provide AI training for small businesses

    OpenAI is launching an initiative to provide hands-on AI training and local support for small business owners across the US. This program aims to help smaller teams use AI to compete with larger organizations by automating routine tasks.

    OpenAI NewsRead the full article

  5. Google Gemini Live adds Guided Vision for real-time visual assistance

    Google's Gemini Live now includes Guided Vision, allowing the AI to 'see' through your camera and provide audio descriptions. Professionals can use this to quickly identify objects, read fine print on documents, or navigate physical environments using real-time AI feedback.

    The Verge AIRead the full article

For builders

4 items

  1. New Codex Cloud enables persistent coding sessions across different devices

    Codex Cloud allows developers to start a task on one machine and resume it on another with all repositories and secrets intact. This facilitates seamless handoffs and mobile reviews for technical teams, though it is primarily a tool for those writing or managing code.

    OpenAI (YouTube)Read the full article

  2. OpenAI showcases the next generation of Codex Originals developers

    OpenAI is highlighting how top developers are using Codex to push the boundaries of what automated software creation can achieve. This signals a shift in how professional developers will build applications, moving from manual coding to higher-level system design.

    OpenAI (YouTube)Read the full article

  3. OpenAI launches Decisions API to manage fast and cheap agentic workflows

    OpenAI has introduced the Decisions API, a new tool designed to handle high-speed automated reasoning tasks. For professionals building custom workflows, this provides a more cost-effective way to deploy autonomous agents that can make quick judgments without the overhead of larger models.

    TechCrunch AIRead the full article

  4. Cost analysis for Claude Code tasks on Opus 5.5

    Anthropic breaks down the actual costs of running specific software engineering tasks using the Claude Code tool on their Opus 5.5 model. This is essential for teams trying to budget for AI-assisted development and understand how different settings impact the final bill.

    Claude Developer BlogRead the full article

Get tomorrow's brief, curated for you.

Subscribe →