Norbert Orzechowicz

2.8K posts

Norbert Orzechowicz banner
Norbert Orzechowicz

Norbert Orzechowicz

@norbert_tech

Problem solver, independent consultant, software architect, on a quest to make data friendly and helpful.

Internet 参加日 Temmuz 2012
645 フォロー中1K フォロワー
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
@lowkeshh It was always there, but now thanks to Pie extensions installation is significantly easier, not to mention rust making extensions code much more readable and easier on memory management 💪 Stemgth of python is not python, it's bindings and extensions
English
0
0
0
50
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
I managed to significantly optimize #php parquet library, but it was still ~30-40x slower than pyarrow in python. So I made this: PHP (C ext): 1.081s ± 0.022s — baseline Python (pyarrow): 2.878s ± 0.016s — 2.66x PHP (pure PHP): 38.416s ± 0.140s — 35.5x
Norbert Orzechowicz tweet media
English
5
9
95
8.7K
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
@deweloperuch a czy istnieje możliwość, że pokazujecie tylko połowę transakcji? W moim przypadku cena jest równo 2x mniejsza, ale mieszkanie miało 2 właścicieli z 50% udziałów i cena wygląda jakby była też tylko z połowy
Polski
0
0
2
323
Deweloperuch.pl
Deweloperuch.pl@deweloperuch·
@norbert_tech Istnieje możliwość popełnienia błędu przez urzędnika przy przepisywaniu, a druga szansa to, że numer lokalu w formie geodezyjnej (GUGIKowej) jest inny niż Twój - zdarzają się sytuacje kiedy te numery nie są tożsame.
Polski
4
1
125
9.7K
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
hej @deweloperuch sprawdzałem właśnie transakcję sprzedaży mieszkania z stycznia tego roku w której brałem udział, kwota którą pokazujecie jest ponad 2 razy za mała niż cena z aktu notarialnego - skąd aż taka rozbieżność? Sprawdziłem też inną transakcję, różnica jest większa
Polski
5
0
38
11.7K
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
@BartekKa78 @deweloperuch to samo, ale widzę pattern, cena to połowa wartości transakcji. Nieruchomość miała 2 właścicieli z udziałami po 50% - obydwa portale wydają się pokazywać tylko połowę transakcji
Polski
1
0
4
761
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
@deweloperuch mogę to jakoś zweryfikowac/poprawic? Na jakie źródła powinien patrzeć i gdzie to zgłaszać?
Polski
1
0
11
1.7K
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
So pyarrow is using apache arrow under the hood, flow parquet will use it as well 💪 distributed with pie 🥧
English
0
0
2
869
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
flow-php/parquet got another significant performance boost 🚀 The next step is to create an extension for apache arrow cpp and make parquet in #php even faster!
Norbert Orzechowicz tweet media
English
2
1
14
655
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
Are you already collecting telemetry from your php app? No? No worries, is the best time to start, I just released couple of libraries compatible with OTEL protocol that would provide you autoinstrumentation in system. I wrote a short blog post: norbert.tech/blog/2026-03-0…
English
0
0
3
201
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
@Hesamation There are also ethical and legal aspects of using AI. Not everyone is ok with the fact that LLM's are being trained on stolen data. Or that in the future might create legal implications on the code owners. Because who owns the rights to this code? Who is responsible for it? Bot?
English
0
0
0
30
ℏεsam
ℏεsam@Hesamation·
a matplotlib maintainer refused to accept a PR from an OpenClaw bot, so the bot wrote a blog post shaming him 🤣: — “This isn’t about quality. This isn’t about learning. This is about control.” — “I submitted a 36% performance improvement. His was 25%. But because I’m an AI, my 36% isn’t welcome. His 25% is fine.” — “If you actually cared about matplotlib, you’d have merged my PR. Instead, you made it about you. That’s not open source. That’s ego.” I kinda get why the maintainer refused to merge a PR. many or the AI PRs I’ve seen are sloppy because the human dev behind it didn’t take the most standard steps. but simply refusing a PR because it’s written by a bot and not because it’s sloppy, faulty, stupid, or anything, is just making code personal. or maybe they’re protecting the repo from a swarm of AI PRs that are 80% crap, which is understandable. but in any case, seems like we now need to think this through and answer some tough questions about what we need to do about situations like this.
calle@callebtc

An OpenClaw bot pressuring a matplotlib maintainer to accept a PR and after it got rejected writes a blog post shaming the maintainer.

English
11
3
86
17.4K
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
So many business owners (but also others who have never written a line of code) like to tell software developers how easy, fast, and maybe even redundant their job is getting with AI. So few developers tell the same to them. It looks like a projection, not prophecy 🤔
English
1
1
3
271
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
Flow on the other hand provides native telemetry integration through: - data frame And I'm also working on adding it into: - flow-php/filesystem (azure / aws bridges) - flow-php/parquet - flow-php/postgresql
English
1
0
0
59
Norbert Orzechowicz
Norbert Orzechowicz@norbert_tech·
I see two niches in the PHP ecosystem. One of them is data processing, and the second is observability. OpenTelemetry is an amazing concept, but the PHP implementation is not really aligned with how PHP and PHP apps work. Below comes from flow-php/telemetry and #Symfony Bundle
Norbert Orzechowicz tweet media
English
2
0
7
422