AI Notes – May 2026

May 2026 felt like a turning point for creative tooling. The biggest story was Google: I/O 2026 made it clear that Gemini is no longer just a chatbot layer but the centre of a broader creative stack spanning multimodal generation, design surfaces, and video workflows. At the same time, AI video stopped feeling like a novelty category and started looking like a mature production layer, with Veo, Runway, Kling, Seedance, and LTX each occupying clearer roles in real workflows. The more subtle shift is that designers are moving up a level: recent 2026 design reporting shows more designers shipping code, building internal tools, and thinking in systems rather than just screens.

Core tools I’d actually use this month

These are the tools that feel most worth paying attention to right now if you work across brand, UI, motion, content, or creative technology.

• Gemini + Google’s creative stack - Google used I/O 2026 to push Gemini 3.5 Flash, Gemini Omni, and a broader “agentic Gemini era,” positioning Gemini as a multimodal layer that can search, generate, edit, and increasingly act across products. For creatives, the important bit is not just another model release; it is that research, design, and video are starting to converge inside one ecosystem.
• Runway - Still one of the strongest choices for cinematic AI video, especially when you care about shot feel, motion quality, and creative control rather than just speed. It remains the tool that feels closest to a film-oriented creative environment rather than a novelty generator.
• Veo + Google Vids - Veo continues to dominate current AI-video usage on at least one major platform, while Vids gives Google a more structured route from prompt or document to finished video inside Workspace. If your work already lives in Docs, Slides, Gmail, and Drive, this stack is becoming hard to ignore.
• Recraft - Recraft remains one of the best tools for designers who need AI outputs that hold up in actual layout, poster, and brand-system contexts, not just moodboard aesthetics. It still feels closer to graphic design than most image generators do.
• Adobe Firefly - Firefly stays highly relevant because it is tightly integrated into Creative Cloud and still one of the safer bets for teams that care about licensing, provenance, and commercially usable outputs. For many agency and in-house teams, it is the most practical AI image layer rather than the most exciting one.
• LTX Studio - LTX increasingly looks like an AI creative operating system rather than just another video tool, combining scripting, storyboarding, shot planning, and assembly inside one environment. That matters because orchestration is becoming more valuable than isolated generation quality.
• Figma AI - Figma’s AI features are now less of a flashy add-on and more of an everyday acceleration layer for product design, documentation, and repetitive interface work. For UI and product workflows, it remains the most natural place for AI to live because it sits directly where design decisions happen.
• Lovable - Lovable remains one of the most compelling vibe-coding tools for designers and creative technologists who want to turn an idea into a working product quickly. It is especially useful when you want to validate a concept or internal tool before involving a full engineering process.
• Ollama + Open Web UI - Local-first AI still matters, especially for anyone working with client documents, sensitive research, or simply wanting more control over models and costs. Ollama plus Open Web UI is still one of the cleanest ways to build a private creative research and ideation stack.
• Canva AI - Canva remains a serious practical tool for fast social graphics, marketing collateral, and mixed-format content when you want one place to move quickly. It is not the most specialist option, but it is increasingly the “good enough and fast enough” tool many teams actually use.

What actually happened in May 2026

Google became a proper creative-platform story

I/O 2026 was the month’s clearest macro event. Google introduced Gemini Omni, described as a model that can create anything from any input, starting with video, and also announced Gemini 3.5 Flash as part of a broader agentic shift. That matters because Google is no longer just offering isolated models; it is designing a connected system where search, generation, editing, and action increasingly sit under the same umbrella.

The creative implication is bigger than a single model benchmark. Google is moving toward tailored, multimodal responses with rich imagery, interactive layouts, narrated video, and dynamic graphics generated on the fly, which starts to blur the line between “search result,” “design output,” and “finished media.” For creative professionals, this suggests that entire micro-experiences, explainers, and content artefacts may soon be generated directly inside search and Workspace-native contexts.

A related move was Pics, Google’s new AI-powered design and image-generation app for Workspace users, positioned around things like invites, social graphics, marketing assets, and mockups. That puts Google much more directly in Canva territory and makes the design layer itself a strategic battleground rather than a side feature of larger AI products.

AI video became a real production layer

By May, the AI-video market looked much more structured. Third-party testing and market analysis kept converging on the same shortlist: Veo 3.1, Runway Gen-4/4.5, Kling, Seedance 2.0, and Pika as the tools that matter right now, each with clearer strengths around quality, controllability, speed, or accessibility. That is a sign of maturity: the conversation is now less “can AI video work?” and more “which engine belongs where in the pipeline?”

Veo remained especially significant because one 2026 video report estimated it held roughly 96% model share on a major AI-video platform, underlining how consolidated parts of the market are becoming. Paired with Google Vids and Gemini, Veo increasingly looks less like a standalone model and more like infrastructure for a broader content ecosystem.

At the same time, platforms like LTX Studio continued pushing the idea that video creation should be managed as a full workflow rather than a sequence of disconnected prompts. For creatives, that is one of the most important shifts of 2026 so far: the winning tool may not be the one with the prettiest clip, but the one that best supports briefs, references, shots, revisions, and exports in one place.

Designers are becoming builders

Recent design-industry research published in May (AI in Design Report 2026) points to a genuine role shift: designers are not just using AI to produce assets, they are increasingly shipping code, building tools, and working closer to product and engineering. Designer Fund’s 2026 report says half of surveyed designers said they had shipped AI-generated code to production, which is an important signal that AI-assisted making is not limited to design engineers anymore.

That fits with the growing relevance of tools like Lovable, v0, Cursor, Replit, and Figma AI in design workflows. It also means design systems matter more, not less: AI-assisted development only works well when teams have solid tokens, components, naming conventions, and documentation for the models to latch onto.

Education and professionalisation kept catching up

May also reinforced the idea that AI for creatives is becoming a formal learning domain. LABASAD promoted its online master focused on generative AI for creatives, with modules on open-source visual tools and applied studio workflows, while mainstream learning platforms like Coursera continued updating creative-AI offerings. That matters because the tool landscape is now dense enough that structured education is becoming more useful than trying to piece everything together from social clips and launch threads.

Stacks that work right now

1. Solo brand designer stack

• Perplexity for market scans and research framing.
• Claude or ChatGPT for naming, messaging, and structured creative briefs.
• Recraft for poster directions, social frames, iconography, and brand-adjacent image work.
• Firefly for commercially safer image generation and Photoshop-native editing.
• Figma AI for layouts, UI directions, and cleaning up repetitive design work.
• Khroma / Fontjoy for quick palette and typography exploration when you need to get unstuck fast.

This stack works because it balances taste, practicality, and handoff. Recraft gives you more design-native outputs than most image models, Firefly keeps the workflow grounded in Creative Cloud, and Figma stays the place where things actually get systematised.

2. AI video stack for social and campaign prototypes

• Gemini / Veo / Google Vids for rapid script-to-video workflows inside a connected ecosystem.
• Runway for more cinematic hero shots and stylised motion tests.
• Kling or Seedance when motion realism, lip-sync, or multi-shot consistency matter more.
• LTX Studio for storyboarding and structuring the whole piece.
• Descript or Riverside for voice, edits, clips, and repurposing.
• HeyGen / Synthesia when avatars, localisation, or presenter-led explainers are part of the brief. This stack reflects the fact that no single video tool does everything best yet. The practical move in May 2026 is to choose one main generation engine, one orchestration layer, and one post-production layer rather than constantly switching between every new model launch.

3. Privacy-first creative technologist stack

• Ollama for local models on a Mac or studio machine.
• Open Web UI as the private interface and RAG layer over internal docs.
• ComfyUI for controllable open image and video pipelines.
• Llama / DeepSeek / other open models depending on hardware and task.
• Cursor or Warp for scripting automations and internal tools.
• Figma / Webflow / Lovable as the outward-facing interface layer.

This stack is still one of the most sensible options for people who care about privacy, client confidentiality, or simply not putting every experiment into a third-party SaaS. It is also increasingly viable because open models and lightweight tooling are improving while the design world becomes more comfortable with AI-assisted building.

Worth trying this month

• Try Gemini Omni with a messy multimodal brief. Feed it a blend of screenshots, notes, audio, and text and see whether Google’s “anything from any input” framing actually reduces iteration time in your workflow.
• Try one real storyboard inside LTX Studio. If you usually jump between Figma, Photoshop, docs, and a video editor, test whether an orchestration-first tool changes how you think about sequencing and approvals.
• Try one internal tool with Lovable or v0. The important question in 2026 is no longer whether designers can code, but how much useful product they can now build without a full engineering cycle.
• Try a local-first research workflow. Run a client brief or research pack through Ollama plus Open Web UI and compare the experience to a cloud-only stack.
• Try using Recraft before Midjourney for one real design task. If the job involves typography, layout logic, or something that has to survive in a brand system, Recraft may save more downstream cleanup than a more visually seductive model.

If you spot any missing links or updates, please DM or comment!