Engineering x product @viktor_com. Prev. 5th at @PostHog. Liberté, Égalité, Fraternité 🇵🇱🇪🇺🏳️🌈
Warsaw
Joined May 2021
- Tweets370
- Following753
- Followers423
- Likes3.2K
"A system that moves steadily is much more stable than one that's frozen and suddenly moves again."
We just got median merge-to-live time in @viktor_com down to 20 min. Only getting this number lower as scale increases ⚡ Fast deployment never mattered so much
I turned the AI image slop on the left into an actual 3D model, on the right.
Or did I? Astra + headless Blender = magic
Here's how to mess with a side project's logo 1/n
One more thing. Could the house start as a sketch, then become half real as it turns? That'd rock.
Tried two variants here. All-sketch and realistic-front. The former fulfilled the goal better, just needed one more iteration for the transition to move diagonally
Aaand here's where I ended up, on the site.
Crazy times, as I've always wanted micro-touches like this in my side projects - just never had the budget in money&time.
Fun? Heck yes. Useful? Only if someone else finds it neat.
Check it out: sketchviz.app
Make
Doors
Great
Again
Panelled doors wasn't a pretty extra but a clever solution to doors expanding with moisture and warping or jamming shut. Because wood expands and contracts in relation to the ambient air humidity (and also rain etc. if outdoors), the wood panels (not nailed or glued) could move freely inside the groves of the frame. This allowed tight doors that shut very close to the door frame, often mere millimeters or fractions of an inch, giving us ncreased privacy and better acoustics.
And they look very charming.
There might be a reason Latvian property is affordable…
But there's a reason I'm upset by this spam:
It's factually incorrect and Latvia is NOT nearly the most affordable in Europe.
Get your spam right
Michael Matloka retweeted
I analysed 255 AI Engineer talks, and extracted 455 ideas presented there.
Evaluation is the biggest topic by far: 98 of 455 ideas (22%). Tools and integrations is next with 51.
Most ideas refine what exists: 37% are optimisations, 22% are about measuring better, 12% add a new capability.
Evidence is thin. 41% of ideas came with no stated result, 35% came with numbers, 6% with a controlled benchmark or A/B test.
I now get the Uber driver's long phone call on a deeper level. A beautifully humanist piece by @NesrineMalik theguardian.com/news/2026/se…
Kudos to @SlackEng for making agents feels at home in Slack! Our guy @viktor_com just graduated from [APP] to [AGENT]
