Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Obscura: The first VPN that can't log your activity(obscura.com)
    discuss
  2. AI Leaders Are Standing on a Liability Landmine(wsj.com)
    discuss
  3. Agent Manager now with mouse support
    discuss
  4. WebMCP Integration at Stripe(twitter.com/stevekaliski)
    discuss
  5. Show HN: Which of your secrets did your coding agent send?(github.com/aggib)
    1comments
  6. ShinyHunters hackers say they breached FBI(reuters.com)
    discuss
  7. Meta testing a 'human concierge' for its new personal AI agent, Muse(reuters.com)
    discuss
  8. Show HN: ClassD-VHF-Transmitter – Covert RF Transmission via Phone Speaker PWM(github.com/ta1eei)
    discuss
  9. The human brain is two separate organs, study finds(stanford.edu)
    discuss
  10. I built the Jev architecture one year ago and open-sourced it(reddit.com)
    discuss
  11. He Said AI? Sharpen the Pitchforks(forta.com)
    discuss
  12. Show HN: junkDrawer.ai – Browse eBay with AI Agent and Dynamic Filtering(junkdrawer.ai)
    discuss
  13. DiscoBox: AI Coding Agents in Disposable Sandboxes(github.com/discobox-ai)
    discuss
  14. Why I'm deciding to go ad-free to support my product(trainmap.co.uk)
    1comments
  15. A Faster Shortest Path Algorithm(vals.ai)
    discuss
  16. New time zone code: PCT for British Columbia(gov.bc.ca)
    1comments
  17. GPT-6 Sol and Luna push the cost-efficiency frontier(artificialanalysis.ai)
    discuss
  18. How Scared Should We Be of A.I.?(nytimes.com)
    discuss
  19. Show HN: Slivingdoc – High scale durable UTF-8 context manager for agents(slivingdoc.dev)
    discuss
  20. A CPU Runs a Program [video](youtube.com)
    discuss
  21. Scientists Have Worked Out When Earth Will Run Out of Oxygen(skyatnightmagazine.com)
    2comments
  22. DeepSeek Elastic Compute:A Sandbox Infrastructure for Effective Agentic Training(arxiv.org)
    discuss
  23. Horowitz Andreessen Academy(twitter.com/gaganbiyani)
    discuss
  24. Google Text: a clean, lightweight editor for Android Desktop (GoogleBook)(play.google.com)
    1comments
  25. How an American naval officer cut weeks off the ocean voyages in the 1840's(twitter.com/bartgonnissen)
    1comments
  26. I Bought a Scanner (No, Really This Time)(leejo.github.io)
    discuss
  27. Trump rejects global entity for AI oversight(politico.com)
    2comments
  28. An autonomous AI company could run itself. It still couldn't find a market(autonomouscompany.substack.com)
    1comments
  29. US to build two military bases in Greenland under new deal with Denmark(bbc.co.uk)
    1comments
  30. Maskr – Catch personal data before it reaches AI chatbots (100% client-side)(chromewebstore.google.com)
    discuss

Show HN: Lightdrift / Image search for AI agents through API and MCP

1 pointsby 51m agolightdrift.ai
0 comments
Greetings HN, I'm Jackson, the founder of Lightdrift. It's an image search API and MCP that lets agents retrieve real images, along with natural language descriptions, metadata, provenance, and license conditions.

I built this because I kept having the same problem and it was pissing me off.

About a year ago, I was vibe coding on edibles. I decided to build a pipeline that generated educational, short form videos about marine biology, because I was watching Planet Earth but I was being distracted by reels, and I needed comfort. I realized video generation struggled with less familiar animals, like the Mola Mola, which is my favorite marine animal because idk there is just something about it that is relatable. Naturally, I decided I would just ground the videos by retrieving images automatically. I realized the existing APIs for this had meh developer experiences, and instantly gave up lol.

Later, I was building an AI education platform. I wanted lessons to include relevant visuals, but generated images weren't consistently reliable for actual species, people, places, etc. Again, I needed image retrieval, and I made a shitty aggregration of APIs and wrappers that ended up becoming the entire project.

I tried Unsplash, Pexels, Brave search, Exa, my own wrappers around Wikimedia Commons. Each helped with part of the problem. Stock libraries worked for broad subjects but didn't consistenly cover the specific things my applications needed. Web search helped locate material, but I still had to resolve the image files, check their sources and usage conditions, and turn inconsistent results into something the application could consume. I wrapped Commons which again left me handling search quality and normalization myself.

The goal of Lightdrift is to bring that tedious ass work into one place. We source, caption, index, and host the images. An app/agent can search in natural language and receive image giles with the context needed to select and use them. The corpus currently contains over two million diverse images, pretty much entirely CC and public domain material.

The license info describes the source's stated permissions/conditions. It isn't a blanket guarantee that every image can be used, but it always gives the information you/your agent need to decide.

I think (know) this could be applied to education tools, presentation builders, and app builders that need relevant imagery without maintaining a separate image pipeline.

I'm currently looking for design partners (but also it's super easy to set up so we don't have to make it official lol). If your application needs particular subjects that we don't cover well, I will work with you to source and index it at no additional cost.

Does image retrieval currently break down in what you're building?

Product: https://lightdrift.ai Docs: https://docs.lightdrift.ai

A quiet thread, for now.Start the conversation on HN ↗