No — my private test didn’t include PFAS or microplastics. I had it done in Aug 2025, when PFAS wasn’t yet part of mandatory Portuguese drinking-water compliance, and those additional tests weren’t really available to me.
The lab checked microbiology + the usual chemical/physical parameters. Separately, Lisbon’s water utility EPAL runs a much broader monitoring program across 1,200+ consumer taps — heavy metals, pesticides, VOCs, etc. They now test PFAS too; the latest result I found was <0.0015 µg/L vs a 0.10 µg/L legal limit.
So my point isn’t that tap water is magically perfect. I’d just rather test the water I actually drink and look at the utility data than assume everyone’s water is toxic 😄 I still filter it anyway.
Могу тоже в Лиссабоне рассказать что и как делаю.
Но я сам для себя пошёл дорогой чуть другой. Я делаю каждый день те вещи которые непрерывно помогают выпускать обновления в продукте быстрее или же упрощают работу как мне так и команде.
Но вот отдать кому-то чтобы кто-то за меня писал промпты и придумывал задачи полностью автономно, пока не дошёл. Скорее всё из-за самих моделей, они ещё не достаточны для меня в этом. И приходиться постоянно следить и направлять. А так в день закрываю обычно по 10 огромных задач, которые в обычной жизни нужно было бы с полгода разработки.
@NotFakeVadim I’ve experimented with fully scripted orchestration.
The problem is agents are good at doing tasks, but still pretty bad at deciding the right next task.
Right now the fastest loop seems to be
AI executing in parallel + human steering the direction.
I think I accidentally found a way to vibe code ~5x faster.
I run 5 Cursor chats in parallel.
While one agent is thinking, I’m reading another response, starting a third task, or reviewing some code.
So basically 5 threads on one machine.
Then I also work from two computers (Mac Studio + laptop) just to keep everything flowing.
And on top of that I’m chatting with a colleague who’s also vibe coding and I’m steering his tasks a bit.
The result is this weird but amazing multi-agent + multi-human flow.
Coding starts to feel less like programming
and more like orchestrating parallel intelligence.
Launched GoClaw — a service around OpenClaw so non-technical users can use it without local setup, deployment, or configs. Telegram integration included. The subscription also includes credits for model inference.
https://t.co/BYpYrbbyVN
MakeUGC V.2 + 7 AI Agents = 550 videos every single day.
Fully realistic ads — cinematic lighting, lifelike motion, perfect pacing — all 100% AI-generated.
Cost: $0
Production time: minutes
Scale: infinite
One autonomous AI engine that creates, tests, and publishes short-form ads — continuously.
It’s live. Campaigns are scaling right now.
Want the full workflow?
Comment + RT “V2” and I’ll DM you the breakdown.
(Must be following.)
@dibuenio Sounds like you've nailed the pain points! What I hate most about trip planning is the endless scrolling through outdated blogs and forums—half the recommendations are from pre-pandemic times, and you end up at a "hidden gem" that's now a chain store. Sign me up for early access
@Nick_Davidov I get the optimism — 5% wins can be outsized. But when you say MIT’s 95% finding is “nuts,” do you mean the data itself is wrong, or just how they interpret it?
@slava_nw Really resonates — especially the idea of funding individuals over startups.
There’s actually a fund by the Liberman brothers doing something along these lines: they invest directly in people, not just ideas or companies.