🤖 AI Weekly Recap (Week 42)

Good morning, AI enthusiast. AI just had another wild week. Google discovered a cancer drug combination that's never been tried in medical history while NVIDIA launched the world's smallest AI supercomputer.

Plus: The most important news and breakthroughs in AI this week.

1. Anthropic Launches “Skills”

Anthropic introduced Skills for Claude, a new feature that lets users teach Claude how to perform specific tasks with custom instructions, scripts, and resources, making it more useful for professional workflows.

→ Skills act as “folders” containing reusable task knowledge, e.g., brand guidelines, Excel automation, or presentation templates
→ Available across Claude, Claude Code, Anthropic’s API, and the Claude Agent SDK
→ Companies like Box, Rakuten, and Canva are already using it to streamline workflows
→ Users can build and share their own Skills to adapt Claude for their organization’s needs

🧰 Who is This Useful For:

• Teams customizing Claude for internal workflows

• Developers building domain-specific AI agents

• Enterprises training Claude on brand or process rules

• Professionals automating daily knowledge work

Try it now → https://claude.ai/new

2. Google’s Gemma AI Model Discovers Potential Cancer Treatment

Google and Yale University researchers announced a major breakthrough: a 27B-parameter Gemma model (C2S-Scale 27B) identified a new way to make “cold” cancer cells visible, a discovery that could reshape cancer diagnosis and treatment.

→ The model analyzed 4,000 drugs and predicted which could act as “conditional amplifiers,” boosting antigen presentation in tumors
→ Identified both known and unexpected compounds, including silmitasertib, which later proved effective in lab validation
→ Combination of interferon + silmitasertib increased antigen presentation, making hidden tumors detectable
→ Marks one of the first real-world biological discoveries led by an AI foundation model

🧰 Who is This Useful For:

• Biomedical researchers exploring AI-driven drug discovery

• Pharma companies accelerating hypothesis generation

• Clinicians focused on earlier cancer detection

• AI scientists pushing the limits of model-based reasoning

Learn More → https://blog.google/technology/ai/google-gemma-ai-cancer-therapy-discovery/

3. Apple Unveils M5 Chip

Apple launched its next-generation M5 chip, powering the new MacBook Pro, iPad Pro, and Vision Pro. The M5 introduces Neural Accelerators in every GPU core, delivering up to 3.5x faster AI performance and setting a new benchmark for on-device intelligence.

→ M5 offers 4x peak compute performance vs. M4, optimized for AI workloads and creative tasks
→ MacBook Pro (from $1,599) and iPad Pro (from $999) feature improved battery life and neural efficiency
→ Vision Pro (from $3,499) adds a Dual Knit Band, visionOS 26, and Apple Immersive apps
→ Devices available for preorder now, shipping October 22 globally

🧰 Who is This Useful For:

• AI developers building on-device inference workflows

• Designers, filmmakers, and creators using intensive GPU applications

• Enterprises prioritizing privacy with on-device AI

• Apple ecosystem users seeking peak AI and creative performance

Learn More → https://www.apple.com/in/newsroom/2025/10/apple-unleashes-m5-the-next-big-leap-in-ai-performance-for-apple-silicon/

4. Nvidia Unveils DGX Spark

Nvidia introduced DGX Spark, the world’s smallest AI supercomputer, a 6×6-inch device delivering 1 petaflop of performance and 128GB of unified memory, all for around $4,000.

→ 99% smaller and 30x cheaper than the original DGX-1 (2016), which cost $129,000
→ Designed for developers, researchers, and edge-AI experimentation
→ Jensen Huang personally delivered a signed DGX Spark to Elon Musk at SpaceX Starbase
→ Preorders open October 15 on Nvidia’s official site

🧰 Who is This Useful For:

• AI researchers needing petaflop-scale power in compact form

• Developers running high-intensity models locally

• Startups and labs exploring affordable supercomputing

• Tech enthusiasts building personal AI clusters

Learn More → https://www.nvidia.com/en-us/products/workstations/dgx-spark/

5. Google Launches Veo 3.1

Google released Veo 3.1, its latest generative video model, now integrated directly into the Flow video editor and Gemini ecosystem. The update enhances realism, sound design, and precise editing control.

→ Adds realistic audio generation and granular edit tools for object blending and motion tracking
→ Users can now add objects seamlessly to match the clip’s visual style; object removal coming soon
→ Supports image-to-video, first-to-last-frame generation, and video extension with new sound support
→ Available across Flow, Gemini App, Vertex AI, and Gemini APIs

🧰 Who is This Useful For:

• Video creators needing fine-tuned AI edits and audio

• Developers integrating video generation into apps

• Brands producing cinematic, on-brand clips with speed

• Filmmakers experimenting with AI-driven storytelling

Try it Now → https://labs.google/fx/tools/flow

6. OpenAI Expands Sora 2

OpenAI’s latest Sora 2 update unlocks new creative power for users. All users can now generate 15-second AI videos, while Pro users can go up to 25 seconds directly on the web.

→ Adds storyboards for Pro users, visual tools for planning and sequencing video ideas
→ Expands video length limits across app and web for deeper storytelling
→ Builds on Sora 2’s lifelike AI “Cameos”, letting users star in custom-generated video scenes

🧰 Who is This Useful For:

• Content creators producing short-form AI videos

• Social media managers testing visual storytelling ideas

• Marketers exploring quick campaign mockups

• Pro users planning multi-scene narratives efficiently

Try it Now → https://sora.chatgpt.com/explore

7. Anthropic Launches Claude Haiku 4.5

Anthropic released Claude Haiku 4.5, a compact AI model designed for speed and cost-efficiency, available to all users. The model outperforms some larger models released just months ago and is ideal for users seeking fast, accurate answers.

→ Smaller and cheaper than Claude Sonnet 4.5 and Opus 4.1
→ Free users get access to Haiku 4.5, while paid users benefit from the most cost-efficient option
→ Can work alongside Sonnet models: Haiku handles subtasks while Sonnet tackles multi-step plans
→ Particularly useful for monitoring high-volume data streams, financial feeds, and fast insights

🧰 Who is This Useful For:

• Businesses running parallel AI workflows for speed and efficiency

• Developers seeking cost-effective, lightweight models

• Mobile and app users wanting quick responses

• Teams managing long-term projects with multi-step AI planning

Try it Now → https://claude.ai/new

8. Runway unveils Apps

Runway unveiled Apps, a collection of specialized AI workflows that make professional editing as simple as writing a text prompt. From removing elements from videos to reshooting products or upscaling footage to 4K, Apps aim to streamline creative workflows.

→ Remove from Video: Upload a video and describe what to remove, no rotoscoping or complex prompting needed
→ Reshoot Product: Transform a single product shot into a library of assets with new settings, styling, or concepts
→ Add Dialogue: Make any subject in an image speak with automatically generated tone, cadence, and delivery

🧰 Who is this useful for:

• Content creators producing high-quality videos quickly

• Marketers creating diverse product assets without photoshoots

• Designers and studios streamlining post-production

• Anyone experimenting with AI-driven video and image workflows

Try it Now → https://runwayml.com/

That's it! See you tomorrow

- Dr. Alvaro Cintas