Nicolas Grilly

Now trying Mastodon. Follow me at @ngrilly@vivaldi.net! #Mastodon #MastodonSocial #mastodonmigration

@halvarflake @bernhardsson Nooooo, that’s even harder to learn than Swedish 😅 If we want to make it really tough, let’s go with Finnish.

@bernhardsson Just standardize on Danish. /me ducks

My newest stupid idea is we should merge the Scandinavian languages. Let’s invent a new one that’s basically the average language. Mandate that it’s used for all public broadcasts. People would learn it in a few months. 25M speakers.

@jonathangrahl @bernhardsson As a French living in Sweden and speaking English, I’d love that 😅

@bernhardsson Better merge the Scandinavian languages with English. Jäs aj äm väry häpy tå spik vitt ju

@mitsuhiko Ignorance is bliss. I won’t look at my own stats.

I data mined my Spotify listening history again and I'm afraid this graph just tells me that I'm turning into an old fart.

@aroussi Spot on

I resisted for a while, but I finally upgraded my Mac to Tahoe (version 26.4). The rumors are true: The adaptation of Liquid Glass to macOS is really ugly, lacking polish, and sometimes even hard to read. First time I've seen @Apple fail this badly at UI and UX.

@notch @ThePrimeagen I use LLM to convert TS to JS so it runs in my browser.

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

So let's start from this post to tell you about my journey with the LLMs and the new Redis data structure I'm implementing right now. Vector Sets were hand-coded, this time I decided to use Claude/Codex as a helpers, and guess what? The work I had to do was huge. Thread:

@excalidraw Small details that matter!

Shipped some improvements around text. First, a single click on a selected text object will put your caret at that exact position.

@excalidraw That’s smart!

Next, it's now easy to spot when a given text has a fixed width. Clicking on the indicator will reset back to auto-resizing.

Famously (there is a beautiful Works in Progress piece on this) in 2016, Geoffrey Hinton told an audience in Toronto that medical schools should stop training radiologists, since AI would soon outperform them at reading scans. Ten years later, there are more radiologists than ever, and they earn more than they did then. Hinton was right about the task, but he was wrong (so far!) on the future of the radiology profession. Times have never been better for them. The gap between those two claims, the difference between tasks and jobs, is the subject of a paper I have written with Jin Li and Yanhui Wu, and that we release today: "Weak Bundle, Strong Bundle: How AI Redraws Job Boundaries." (Very relatedly we are also finishing the first draft of our book "Messy Jobs" on AI and Jobs!! You will be the first to hear). We start from the observation that the growing literature on AI and labor markets measures the AI shock by task exposure: people count how many tasks AI can perform in a given occupation AI can perform, and infer that more exposure means more displacement. Eloundou et al. published a paper in Science in 2024 that started this literature, and many follow the same logic. The inference they make is that the more exposed tasks, the worse the outcomes. This is incomplete, because labor markets price jobs, not tasks. A radiologist does not just sell image classification, but does many other jobs: triages cases, communicates with other physicians, trains residents, makes the difficult decisions, and signs a diagnosis. The market buys a bundled service. The question AI poses is not whether it can do one task inside the bundle. The question is whether that task can be pulled out. Thread (1/3) dropbox.com/scl/fo/689u1g7…

@michuk @paddi_hansen @alex @euinc_petition It’s far from perfect indeed, but I’d rather use this than the mess we have today. That’s still a step in the right direction. It’s not like the USA are perfectly harmonized either: most startups incorporate in Delaware because the rules are different in every state.

@paddi_hansen @alex Unfortunately this proposal has little to do with what founders and investors asked for when lobbying for @euinc_petition. All we need is standardization while the proposal suggests there will be 27 different implementations by each member state. Nobody will use it :(

The new EU Inc - how it works and what's still missing 🇪🇺 3 days ago, the EU tabled its long-discussed proposal for a “28th regime” — a new optional company form: EU Inc. → Europe still has 27 different company law systems → A major barrier for startups scaling across the Single Market → A core theme of the Draghi & Letta reports and the EU competitiveness agenda What is EU Inc.? → An optional EU-wide corporate form → A single rulebook you can opt into instead of national regimes → Still anchored in a Member State (not a fully separate EU system) Registration / setup → Fully digital (no paper, no physical presence) → Via EU interface (BRIS) or national registers → 48h incorporation (with templates), max €100 → 5 working days with custom statutes → “Once-only” principle: data reused across tax, VAT, etc. Governance → Board structure (at least 1 EU-resident director) → Fully digital governance (online meetings, e-voting) → Harmonised duties, liability, conflict rules → Minority shareholder protections (incl. exit rights) Shares & fundraising (key for startups) → Fully digital shares (no paper, no notaries required) → Free transfer of shares across borders → Flexible share classes (incl. multiple voting rights) → No minimum capital (can be €0) → No nominal value shares by default → Designed to support SAFEs, convertibles, warrants Employee stock options (big one) → New EU-wide ESO regime → Taxation only at sale of shares (not grant/vesting/exercise) → Aimed at fixing Europe’s “dry tax” problem Cross-border scaling → Branches & subsidiaries set up digitally EU-wide → Company data shared automatically via BRIS → EU Company Certificate + digital power of attorney → No re-submission of the same info across countries So what’s the catch? (key criticism from the startups) → Core areas still national: tax, labour, insolvency -> this won't change anytime soon → Not a fully unified EU company code → Disputes in countries remain with national courts (no EU-level system) -> probably the main point of criticism at the moment What’s next? → Parliament + Member States negotiations starting now → Political deal targeted by end of 2026 → Likely application ~2028 (after entry into force + 12 months) Bottom line: One of the most ambitious EU steps in history toward a real Single Market for companies. Not yet “one rulebook” across the board - but a big step in the right direction. Pushback from usual interest groups will be significant, so outcome and final shape remain tbd. The history of similar proposals such as the Societas Europaea in 2004 shows how tough it is to crack that nut. But Fingers crossed. Congrats for this initial success to long-time supporters like @euinc_petition @euacc @euacchq. Keep pushing!

@thdxr I don’t get it either. I also see people who push code produced by their agent to review in GitHub.

i still don't get why we need to push code up to get an LLM review via awkward github ui hacks opencode has /review which can also do things like run your code to check things but a full time team focused on this would do it better, i just don't like the workflow they offer

@paulg That’s also why Axios articles use bulleted lists…

The fact that AIs tend to answer you in bulleted lists tells us something important, though somewhat depressing: people can't read. They don't do this by accident. What you're seeing is an implicit portrait of the median user.

Love this. The NY Times of coding agents.

the CEO of htmx and the CEO of Vercel need to talk

@jokull Agent loops don’t need React. They need fast compile, deterministic UI, no bundle drama. Go + htmx + templ + templui is basically AI-native web infra. templui.io

Kagi Translate now supports LinkedIn Speak as an output language.