← Back to all sparks
I

Icons8

DESIGN
Velocity2.5

Design resources platform offering icons, illustrations, photos, and AI-powered design tools.

Icons8 writes about AI tools it doesn't sell, and occasionally ships one it does.

design assetsgenerative aitool benchmarksgrounded generation
Current state
The feed publishes sporadically — roughly monthly — and mostly tests other companies' AI tools: upscalers benchmarked on a real print banner, mockup generators, video models. The newest post argues about which 2027 design trends survive once anyone can generate them. The exception in the window is a build post for a website generator that takes Google Maps reviews as its only input.
Where it's heading
The editorial position is a design-asset company reckoning with generation making assets cheap, and the answer it has shipped is grounding: a generator constrained to real customer reviews rather than free-form output. Publishing is too infrequent to read a roadmap from.
Prediction
Expect more comparison testing of generative tools, with occasional build posts when Icons8 ships something of its own; cadence makes timing unpredictable.

Recent moves

  1. 6d ago

    Design trends for 2027

    A trends piece on which design directions survive commoditization by generators — editorial positioning, not a product change.

    View source ↗
  2. 1mo ago

    Best AI Image Upscaler for Printing in 2026: I Tested 6 Models on a Real Banner

    A hands-on benchmark of six upscalers on a print file.

    View source ↗
  3. 2mo ago

    Best AI mockup generators

    A roundup of AI mockup tools.

    View source ↗
  4. 3mo ago

    We built a website generator that can’t hallucinate

    The one entry describing something Icons8 built: a website generator restricted to Google Maps review text, which is the grounding answer to the trends argument the feed keeps making.

    View source ↗
  5. 3mo ago

    Popular symbols breakdown

    A design-history piece on common symbols.

    View source ↗
  6. 4mo ago

    Best Video Generation Models in 2026

    A comparison of video generation models.

    View source ↗