<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>Latest from Nico Martin</title>
    <link>https://nico.dev/latest/</link>
    <description>Projects, articles, videos, pictures and podcast appearances from Nico Martin.</description>
    <language>en</language>
    <lastBuildDate>Sun, 16 Aug 2026 05:12:58 GMT</lastBuildDate>
    <atom:link xmlns:atom="http://www.w3.org/2005/Atom" href="https://nico.dev/latest/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>LFM2.5 Edge Agent</title>
      <link>https://huggingface.co/spaces/LiquidAI/LFM2.5-2.6B-WebGPU</link>
      <guid isPermaLink="false">project:2026-08-05:LFM2.5 Edge Agent</guid>
      <pubDate>Wed, 05 Aug 2026 12:00:00 GMT</pubDate>
      <category>project</category>
      <description>I built a research agent that runs entirely in the browser and puts its full think, plan, and act loop on display.

Give it a question and it:

- Creates an explicit action plan
- Reasons about the next step
- Chooses and calls tools
- Inspects each result
- Updates the plan and repeats

It does not stop after one tool call. The agent keeps looping until every step in the plan is complete.

Along the way, it can ask clarifying questions, use local context, delegate focused research to an isolated Wikipedia subagent, and collect sources. Once the plan is done, it turns the evidence into a cited response and downloadable research paper.

The incredible part is the size: just 2.6B parameters.

That makes LFM2.5 seriously fast on-device with WebGPU, while still being capable enough to plan, reason, use tools, and finish multi-step objectives. The model, agent loop, and inference pipeline all run locally. No inference server, no account, and prompts and outputs stay in the browser.

Small models are not just autocomplete. They can be agents at the edge.</description>
    </item>
    <item>
      <title>After using windows basically my whole life, it was on-device AI performance tha...</title>
      <link>https://www.instagram.com/p/DbDQMAKs_yl/</link>
      <guid isPermaLink="false">instagram:2026-07-21:After using windows basically my whole life, it was on-device AI performance tha...</guid>
      <pubDate>Tue, 21 Jul 2026 12:00:00 GMT</pubDate>
      <category>instagram</category>
      <description>After using windows basically my whole life, it was on-device AI performance that finally pulled me over to apple. And honestly? Never regretted it for a second.

I bought this M3 MBP in feb 2024. 2.5 years later I&apos;ve never owned a laptop this long and there&apos;s still no end in sight. Runs like day one, every single day.

The stickers tell the story 🚀

What&apos;s the longest you&apos;ve kept a laptop?</description>
    </item>
    <item>
      <title>Transformers.js mit Nico Martin</title>
      <link>https://www.programmier.bar/podcast/deep-dive-211-transformers-js-mit-nico-martin</link>
      <guid isPermaLink="false">podcast:2026-07-07:Transformers.js mit Nico Martin</guid>
      <pubDate>Tue, 07 Jul 2026 12:00:00 GMT</pubDate>
      <category>podcast</category>
      <description>In dieser Folge begrüßen wir einen alten Bekannten in der programmier.bar: Nico Martin, Machine Learning Engineer bei Hugging Face und gern gesehener Podcast-Gast und programmier.con-Speaker.</description>
    </item>
    <item>
      <title>After my week started in Bucharest, it ended at a rock concert in Oslo. Sort of....</title>
      <link>https://www.instagram.com/p/DZPjY3qDFIJ/</link>
      <guid isPermaLink="false">instagram:2026-06-06:After my week started in Bucharest, it ended at a rock concert in Oslo. Sort of....</guid>
      <pubDate>Sat, 06 Jun 2026 12:00:00 GMT</pubDate>
      <category>instagram</category>
      <description>After my week started in Bucharest, it ended at a rock concert in Oslo. Sort of.

#ReactNorway took place at Rockefeller, a legendary Oslo concert hall that&apos;s hosted everything from Norwegian metal festivals to international headliners. And they leaned fully into it. The stage, the lighting, the crowd energy: this wasn&apos;t a conference room with a projector. It felt like a show.

One track, no escape, full attention. It&apos;s my favourite conference format, full stop. Everyone shares the same experience, hallway conversations are actually about the talks, and there&apos;s zero FOMO.

I spoke about my in-browser Jarvis demo, which was very well received. And it sparked some great discussions how far on-device AI has already come! Always a good sign.

My time in Oslo was short but sweet. Just long enough to soak up the city a little (and the surrounding nature). One thing I was not prepared for: sunrise at 4am, sunset at 11pm. My body still hasn&apos;t recovered.

Thanks for the warm welcome, Oslo, and huge thanks to Aina, Nebojsa and the whole Team🤘</description>
    </item>
    <item>
      <title>CEE&apos;s biggest tech conference... ...also: Bucharest is great. This week I had th...</title>
      <link>https://www.instagram.com/p/DZJ5p7ejFkG/</link>
      <guid isPermaLink="false">instagram:2026-06-04:CEE&apos;s biggest tech conference... ...also: Bucharest is great. This week I had th...</guid>
      <pubDate>Thu, 04 Jun 2026 12:00:00 GMT</pubDate>
      <category>instagram</category>
      <description>CEE&apos;s biggest tech conference...
...also: Bucharest is great.

This week I had the pleasure of speaking at @devtalksro, the largest tech conference in Central and Eastern Europe, and honestly, it showed.

The expo floor alone was something else: from Google all the way to students showcasing their own robot projects. That range of scale, from industry giants to the next generation of builders, is exactly what a healthy tech community looks like.

I didn&apos;t get to see as many talks as I would have liked: my morning was taken up by my own session, where I walked through the story behind my in-browser Jarvis and the ideas that shaped it. In the afternoon I got to go even deeper in a roundtable workshop, diving hands-on into the agent loop with the participants.

Both were a blast.

What I didn&apos;t expect was how much Bucharest had to offer. A charming old town, the striking architecture of the communist era, and a very much alive modern city vibe. Definitely underrated.

Great connections, great energy, great city. Thanks for having me, DevTalksRomania 🙌</description>
    </item>
    <item>
      <title>Google I/O 2026 aus Mountain View</title>
      <link>https://www.programmier.bar/podcast/spezialfolge-google-i-o-2026-aus-mountain-view</link>
      <guid isPermaLink="false">podcast:2026-06-02:Google I/O 2026 aus Mountain View</guid>
      <pubDate>Tue, 02 Jun 2026 12:00:00 GMT</pubDate>
      <category>podcast</category>
      <description>Die Google I/O ist für viele Entwickler:innen vor allem durch ihre Ankündigungen bekannt – doch wie fühlt sich die Konferenz eigentlich vor Ort an? Darüber sprechen Jan und Dennis mit Nico Martin, der in diesem Jahr nach Mountain View gereist ist.

Nico berichtet von seinen Eindrücken auf dem weitläufigen Google-Gelände, den großen Keynotes und den zahlreichen Möglichkeiten, direkt mit den Teams hinter Chrome, Gemini und anderen Google-Produkten ins Gespräch zu kommen.

Natürlich geht es auch um die großen Themen der diesjährigen Google I/O. Die Runde diskutiert die aktuellen Gemini-Ankündigungen, den hohen Produktionsaufwand hinter den Keynotes und die Frage, wie viel davon wirklich neu war. Besonders spannend waren für Nico die Produkt-Demos vor Ort, darunter die neue Google-Brille, die er selbst ausprobieren konnte.

Außerdem werfen Jan, Dennis und Nico einen Blick auf die Rolle von Entwickler:innenkonferenzen im Allgemeinen. Sie vergleichen die Google I/O mit Formaten wie der Google Cloud Next und den lokalen Google I/O Connect Events und diskutieren, für wen sich ein Besuch lohnt.</description>
    </item>
    <item>
      <title>How Transformers.js Works: AI Models in JavaScript, Explained</title>
      <link>https://www.youtube.com/watch?v=A_6UOxQSvvY</link>
      <guid isPermaLink="false">video:2026-05-27:How Transformers.js Works: AI Models in JavaScript, Explained</guid>
      <pubDate>Wed, 27 May 2026 12:00:00 GMT</pubDate>
      <category>video</category>
      <description>Transformers.js brings state-of-the-art machine learning to JavaScriprt.

In this intro, I explain the mental model behind Transformers.js: tensors, neural networks, ONNX, quantization, model loading, preprocessing, postprocessing, and the pipeline() API.

We also look at what happens under the hood for text-generation and depth-estimation, and why one high-level API can support many different AI tasks.
</description>
    </item>
    <item>
      <title>Back from my first Silicon Valley trip. A lot to unpack. First of all: a great c...</title>
      <link>https://www.instagram.com/p/DYrhHPAjNZf/</link>
      <guid isPermaLink="false">instagram:2026-05-23:Back from my first Silicon Valley trip. A lot to unpack. First of all: a great c...</guid>
      <pubDate>Sat, 23 May 2026 12:00:00 GMT</pubDate>
      <category>instagram</category>
      <description>Back from my first Silicon Valley trip. A lot to unpack.

First of all: a great chance for Joshua Lochner and me to have a little Transformers.js offsite. We&apos;d met in Amsterdam almost exactly a year ago, but that was before I joined Hugging Face. Remote work is great, but meeting in person every now and then? Priceless.

𝗗𝗮𝘆𝘀 𝟭 𝗮𝗻𝗱 𝟮: 𝗚𝗼𝗼𝗴𝗹𝗲 𝗜/𝗢
Finally meeting so many internet friends in person. Years of comments, DMs, and GitHub threads and suddenly you&apos;re grabbing coffee with them. That alone was worth the flight..

On the tech side: lots of Gemini in everything. More deep integration than groundbreaking innovation, honestly. The exception: Gemini Omni. All-modal in, all-modal out. That one gave me pause.

Personal highlight: the Gemma 4 sessions, including our own live demo of a Reachy Mini robot powered by Transformers.js and Gemma 4. Seeing on-device WebML power a robot in real time is exactly why I do this work 😊

Dinner with the LiteRT.js team was one of the best conversations of the trip. On paper, Google&apos;s web ML inference library is a competitor to Transformers.js. In practice? Not even close to competitive in spirit. We all want the same thing: make AI inference in the browser better. Rising tide lifts all boats.

𝗗𝗮𝘆 𝟯: 𝗦𝗮𝗻 𝗙𝗿𝗮𝗻𝗰𝗶𝘀𝗰𝗼
The city genuinely surprised me. Stunning scenery, Golden Gate, and the Google I/O After Hours event in the evening.. great connections made.

𝗢𝗻𝗲 𝗺𝗼𝗿𝗲 𝘁𝗵𝗶𝗻𝗴: 𝗪𝗮𝘆𝗺𝗼
In Europe, self-driving cars still feel distant. In the Bay Area, they&apos;re just... normal. Smooth, safe, easy, whether on an open road or in rush hour. The future is already here.

10/10. Can&apos;t wait to come back 😊</description>
    </item>
    <item>
      <title>Biggest learning from @uphill_conf &apos;26: being a speaker is so much easier than b...</title>
      <link>https://www.instagram.com/p/DYUGe5ODL5D/</link>
      <guid isPermaLink="false">instagram:2026-05-14:Biggest learning from @uphill_conf &apos;26: being a speaker is so much easier than b...</guid>
      <pubDate>Thu, 14 May 2026 12:00:00 GMT</pubDate>
      <category>instagram</category>
      <description>Biggest learning from @uphill_conf &apos;26: being a speaker is so much easier than being an organizer 😅

After the last two years on stage, I switched sides this year. Nine months of prep, a lot of group chats, and one very humbling ropes course at the speaker event in Grindelwald later, we pulled it off!

over 300 attendees enjoyed a sunny and inspiring day on the Gurten. This year with a second stage and a focus on valuable insights, optimized solutions, and lessons learned from running AI apps in production to share with the developer community.

A few highlights from the talks:

Imanol Schlag gave us a behind-the-scenes look at Apertus, a fully open research project from Switzerland. can&apos;t wait for 1.5.
Omar Sanseviero walked through what Gemma4 means for on-device AI. New benchmarks, solid technical depth.
@charlie__g closed the day with creative and experimental AI interfaces, demoed live on stage. perfect way to end a conference.

And my on-stage-contribution this year? I got to moderate the open source panel with Ben Burtenshaw, @jonnyburger, Panos Vagenas and Romain Jourdan.
Three things stuck with me:
💡 On OSS business models: AI makes everything grow exponentially. Without a scalable model underneath, that growth becomes a problem.
💡 On OSS security: security by obscurity was always the wrong approach. The more interesting question is how to leverage AI for security itself.
💡 On OSS contribution overload: Large OSS projects are drowning in PRs, most of them one-prompt, no real thought behind them. One reframe that landed: stop treating PRs as contributions and start treating them as signals. A spike of PRs around one topic might point to a real solution worth exploring.

Thanks to everyone who joined us on the Gurten. see you next year 🏔️

And a huge thanks to Matthias, Christian, Johann, Roland, Christof, Marco and Tobias for organizing this with me.

#UphillConf #OpenSource #AI #Conference</description>
    </item>
    <item>
      <title>Browser ML at Scale with Nico Martin</title>
      <link>https://neciudan.dev/takeaways/browser-ml-at-scale-with-nico-martin</link>
      <guid isPermaLink="false">podcast:2026-04-26:Browser ML at Scale with Nico Martin</guid>
      <pubDate>Sun, 26 Apr 2026 12:00:00 GMT</pubDate>
      <category>podcast</category>
      <description>This week, I spoke with Nico Martin, open source ML engineer at Hugging Face working on Transformers.js, and a Google Developer Expert in AI and web technology. Nico&apos;s path is unusual — he studied to become a teacher, worked as a ski instructor and a windsurf instructor, taught himself web development by building sites for friends, and went freelance for years before joining Hugging Face about six months ago. The role pulls together his obsession with progressive web apps, on-device AI, and Transformers.js — a library he was already building demos with before version one even shipped.

In this episode, we cover how Transformers.js actually runs models in the browser, what an ML model really is under the hood, how tokenizers and quantization work, why ONNX matters as a runtime-agnostic standard, where WebGPU fits next to WebAssembly, and what the Prompt API in Chrome means for the future of in-browser LLMs.</description>
    </item>
    <item>
      <title>Gemma 4 Browser Assistant</title>
      <link>https://chromewebstore.google.com/detail/transformersjs-gemma-4-br/dhaknnnkcdkjhcclchmnfdhddoehoool</link>
      <guid isPermaLink="false">project:2026-04-23:Gemma 4 Browser Assistant</guid>
      <pubDate>Thu, 23 Apr 2026 12:00:00 GMT</pubDate>
      <category>project</category>
      <description>An on-device AI assistant that runs entirely in your browser using WebGPU and Transformers.js. This Chrome extension provides an intelligent agent that can understand natural language commands and interact with your browser through a set of specialized tools.</description>
    </item>
    <item>
      <title>How to Use Transformers.js in a Chrome Extension</title>
      <link>https://huggingface.co/blog/transformersjs-chrome-extension</link>
      <guid isPermaLink="false">blogpost:2026-04-23:How to Use Transformers.js in a Chrome Extension</guid>
      <pubDate>Thu, 23 Apr 2026 12:00:00 GMT</pubDate>
      <category>blogpost</category>
      <description>An article I wrote about how to build a Chrome extension with on-device AI using Transformers.js and Gemma 4, covering Manifest V3 architecture, model loading in service workers, and building an agentic tool-calling loop.</description>
    </item>
    <item>
      <title>Transformers.js v4: State-of-the-art machine learning for the web</title>
      <link>https://youtu.be/KnhppkY4gHs</link>
      <guid isPermaLink="false">video:2026-03-30:Transformers.js v4: State-of-the-art machine learning for the web</guid>
      <pubDate>Mon, 30 Mar 2026 12:00:00 GMT</pubDate>
      <category>video</category>
      <description>Transformers.js v4 is out!
</description>
    </item>
    <item>
      <title>JARVIS</title>
      <link>https://jarvis.nico.dev/</link>
      <guid isPermaLink="false">project:2025-08-27:JARVIS</guid>
      <pubDate>Wed, 27 Aug 2025 12:00:00 GMT</pubDate>
      <category>project</category>
      <description>A modern AI assistant web application featuring voice interaction, real-time conversation, and MCP (Model Context Protocol) server integration.</description>
    </item>
    <item>
      <title>I&apos;m joining 🤗 Hugging Face</title>
      <link>https://www.linkedin.com/feed/update/urn:li:activity:7352927768600354816/</link>
      <guid isPermaLink="false">blogpost:2025-07-21:I&apos;m joining 🤗 Hugging Face</guid>
      <pubDate>Mon, 21 Jul 2025 12:00:00 GMT</pubDate>
      <category>blogpost</category>
      <description>Beyond happy to announce that I&apos;m joining 🤗 Hugging Face as a Machine Learning Engineer focused on WebML!</description>
    </item>
    <item>
      <title>PWAs mit Nico Martin</title>
      <link>https://www.programmier.bar/podcast/deep-dive-154-pwas-mit-nico-martin</link>
      <guid isPermaLink="false">podcast:2024-07-26:PWAs mit Nico Martin</guid>
      <pubDate>Fri, 26 Jul 2024 12:00:00 GMT</pubDate>
      <category>podcast</category>
      <description>Lange war es still um PWAs, doch jetzt ist das Thema Progressive Web Apps gerade aktueller denn je: Zum einen hat der Bereich viel Aufmerksamkeit erfahren, als Apple im Frühjahr 2024 kurz davor war, das Feature unter iOS in Europa dauerhaft abzuschalten. Zum anderen erleben PWAs dank der neuen WebGPU-APIs und dem Hype um AI eine Art Renaissance.

Aber was sind eigentlich Progressive Web Apps und was unterscheidet sie von „normalen“ Webseiten oder Bookmarks auf dem Homescreen? Und macht die Art der Anwendung für Endnutzer:innen wirklich einen Unterschied?

All diesen Fragen und vielen weiteren gehen Dennis und Jan nach und sprechen dafür mit Nico Martin, Google Developer Expert fürs Web. Dabei geht’s nicht nur um APIs, Browser-Implementierungen und Webstandards, sondern auch darum, was User:innen von Webseiten erwarten und welche UX für sie relevant ist.</description>
    </item>
    <item>
      <title>Ask my PDF</title>
      <link>https://pdf.nico.dev/</link>
      <guid isPermaLink="false">project:2024-06-04:Ask my PDF</guid>
      <pubDate>Tue, 04 Jun 2024 12:00:00 GMT</pubDate>
      <category>project</category>
      <description>A Webapp that uses Retrieval Augmented Generation (RAG) and Large Language Models to interact with a PDF directly in the browser.</description>
    </item>
    <item>
      <title>Git Installer</title>
      <link>https://www.git-installer.com/</link>
      <guid isPermaLink="false">project:2022-11-22:Git Installer</guid>
      <pubDate>Tue, 22 Nov 2022 12:00:00 GMT</pubDate>
      <category>project</category>
      <description>Install and update WordPress themes and plugins directly from your Git repository via GitHub, Gitlab or Bitbucket.</description>
    </item>
    <item>
      <title>SpeedWheels BLE Car</title>
      <link>https://speed-wheels.nico.dev/</link>
      <guid isPermaLink="false">project:2022-03-30:SpeedWheels BLE Car</guid>
      <pubDate>Wed, 30 Mar 2022 12:00:00 GMT</pubDate>
      <category>project</category>
      <description>A toy car based on an Arduino Nano RP2040 that communicates over Bluetooth Low Energy and can be cntrolled with a webapplication that uses the Web Bluetooth API.</description>
    </item>
    <item>
      <title>Revision 521: WebBluetooth, WebUSB und WebNFC</title>
      <link>https://workingdraft.de/521/</link>
      <guid isPermaLink="false">podcast:2022-03-16:Revision 521: WebBluetooth, WebUSB und WebNFC</guid>
      <pubDate>Wed, 16 Mar 2022 12:00:00 GMT</pubDate>
      <category>podcast</category>
      <description>Nico Martin, Google Developer Expert und Webtech-Nerd, erklärt in dieser Revision Schepp und Peter alles zu WebBluetooth, WebUSB und WebNFC!</description>
    </item>
    <item>
      <title>[md.edit]</title>
      <link>https://md.nico.dev/</link>
      <guid isPermaLink="false">project:2021-12-17:[md.edit]</guid>
      <pubDate>Fri, 17 Dec 2021 12:00:00 GMT</pubDate>
      <category>project</category>
      <description>A web based markdown editor that uses modern progressive web app features to provide a great, cross-platform editing experience. The main goal of this project is to showcase the File System Access API. This API allows a web app to access local files. That means files can be opened and saved directly in the browser.</description>
    </item>
    <item>
      <title>Share Target API</title>
      <link>https://dev.to/nicomartin/share-target-api-5263</link>
      <guid isPermaLink="false">blogpost:2020-05-03:Share Target API</guid>
      <pubDate>Sun, 03 May 2020 12:00:00 GMT</pubDate>
      <category>blogpost</category>
      <description>I think it&apos;s fair to say that sharing is one of the most time-saving interactions on our phones. So I was more than thrilled when I&apos;ve heard about the Share Target API.</description>
    </item>
    <item>
      <title>Creating Scheduled Push Notifications</title>
      <link>https://css-tricks.com/creating-scheduled-push-notifications/</link>
      <guid isPermaLink="false">blogpost:2020-04-10:Creating Scheduled Push Notifications</guid>
      <pubDate>Fri, 10 Apr 2020 12:00:00 GMT</pubDate>
      <category>blogpost</category>
      <description>Scheduled is the key word there — that’s a fairly new thing! When a push notification is scheduled (i.e. &quot;Take your pill&quot; or &quot;You&apos;ve got a flight in 3 hours&quot;) that means it can be...</description>
    </item>
  </channel>
</rss>