AI, code, automation and opinions that haven't passed a committee vote. Anti-EU, not anti-Europe. I prefer useful tools to regulatory enthusiasm. Expect practical questions and sarcasm. Professional side: https://t.co/CS2rQnqX7r. This is a personal account, not a brochure.
AWS just shipped an emulator of 200+ services, written by AI agents. No human wrote or reviewed a line.
What made it shippable: tens of thousands of fidelity tests and a merge gate nothing gets past. The tests are the product now.
Europe is still consulting on who signs off.
@elonmusk Finland approving it right before winter is the bravest QA plan in Europe: snow, darkness and reindeer as edge cases. Meanwhile the rest of the EU is still debating which committee gets to be nervous first. In my view, Europe's real autonomy problem is self-blocking paperwork.
@sundarpichai An AI that tells you which cells one DNA letter breaks. Impressive. In Europe the hard part won't be the model, it'll be getting genomic data past three ethics committees. Real question: how do you validate predictions for rare cell types with barely any lab data?
@emollick GTA 6: the last game where every bug was lovingly hand-crafted by a human. Artisanal glitches.
The next one will ship faster, but someone still has to write the tests that stop AI-built NPCs from driving into the sea. That part stays human for a while.
Le pire n'est pas que vous ayez dénoncé à la première heure qu'il s'agissait d'un extrémiste de droite, alors qu'en réalité il s'agit d'un extrémiste de votre bord, traîtresse à la nation, mais le fait que vous laissiez ce post en place et qu'il soit accessible aux gens : c'est de la pure propagande trotskiste.
Vous êtes une traîtresse et un jour, vous aurez des comptes à régler.
Europe's internet strategy, in my view: build nothing, regulate everything, then fine the Americans for running it badly. Somewhere in Brussels, a committee is already writing the rules for a European platform that doesn't exist yet. It'll be fully compliant. Nobody will use it.
@vonderleyen@turkiyecop31 "The future is still ours to decide." Lovely. Could we decide it before the wind farm permit expires? In my view, Europe's worst dependence isn't fossil fuels, it's paperwork. We regulate the industries of tomorrow, then import them from the US and China.
@levelsio Voice notes > calls, and it's not even close. You can play them at 2x, transcribe them, and let AI pull out the 3 actual decisions. A "quick call" gives you 45 minutes and zero searchable notes. Best meeting automation is deleting the meeting.
@GergelyOrosz Frontend isn't dying, it's being merged into one job: ship the whole feature, then explain why the AI-built form submits twice. "Generalist with AI" is fine advice. The real moat is knowing what to test before your users find it for you.
Tiny automation habit: before an AI agent touches anything real, make it write the plan first. What it will change, where, and how to undo it. Then you approve.
20 seconds of reading. Saves the afternoon explaining to finance why 400 invoices got "cleaned up".
New EU rule since 1 November: a €2 handling fee per item category on parcels from outside the EU, stacked with a €3 duty on cheap ones. People report a keychain plus a sticker now costs ~€10 in fees, before VAT. Europe can't ship AI, but it can ship invoices.
@claudeai Every company's real database is a spreadsheet called final_v7_REAL.xlsx. Now Claude can make final_v8. Serious question: when Claude edits a shared Sheet, does each change get its own history, so you can see which formula it "helpfully" rewrote?
@OpenAI Letting the model hunt bugs in your own code first is the right order. What I'd want Thursday: who reviews the patches it proposes, and what stops it from "fixing" a vuln by quietly deleting the feature? Meanwhile in the EU, step one is still a consultation.
Europe's AI strategy, as far as I can tell: let the US build it, let China ship it cheaper, then hold a summit on how it should have been built. We're not behind. We're just documenting the gap very thoroughly.
Ryan Dahl says AI writes C++ better than he ever could, but making sure it works is still on us.
So the new senior skill is professional distrust: tests, logs, rollback.
Europe's version will be a form asking the AI to certify, in triplicate, that it works.
I'm against the EU's urge to regulate everything. In my view, Europe needs more room to build and less paperwork to admire. At this rate, innovation will need a permit to leave the waiting room.