Ignacy

[Ignacy @ignacy130]

@KacperPlazynski @MaciekAntczak "Zgodnie z tą propozycją, kwota z budżetu państwa przeznaczona na dopłaty do BK2% ma wzrosnąć z ok. 11,3 mld zł obecnie do niemal 16,1 mld zł w latach 2024-2032."

https://t.co/l35TD4Dyxl

[Ignacy @ignacy130]

@Evgastap @Gregorein So big it's down

[Ignacy @ignacy130]

@0xlelouch_ Stopped reading at mechanical keyboard

[Scott Hanselman 🌮 @shanselman]

I’m seeing some hot takes that AI assisted coding means that you don’t have to be technical anymore. That’s only gonna last you until the first database migration, or the first security issue, or the first cloud migration, or the first scale out, or the first major regression, or the first refactor that ends in slop. I am finding that I’m learning more and I have to be as technical or more technical than ever before to get the kinds of high-quality output that I expect of any code, regardless of whether it comes from my fingertips or someone else’s - including an AI.

Whether your source comes from open source libraries, your own hands, or an AI via your clever prompt, there is exactly one responsible person for the output. That is you.

I never want to be accused of gatekeeping AI assisted programming, as non-technical people can get a lot of interesting work done. Until they hit a wall, and it’s gonna surprise them how quickly they either need to get technical, or get a technical person to help untangle the mess they’ve made.

The art and science of programming is taking intent and turning it into shipping products. I will never blame an AI - nor should you - for bad output. Own the code that you ship.

[Ignacy @ignacy130]

@ellobosalvaje It was mediocre before, it's mediocre after.

[Ignacy @ignacy130]

Last week I've read a LinkedIn post where a guy is positively surprised by creatijg a mobile friendly Power BI dashboard in under a day. Something I did first time 7 years ago. It's still early!

[akano @princessakano]

God, I love this genre of "Cloud Thought Leader" takes. It must be so peaceful up there in the ivory tower of Solutions Architecture diagrams where the boxes always connect neatly to the cylinders.

>Most apps are toys.
>Perf isn't even measured.

I beg you. Please. Give me the toys.

I would kill to manage an infrastructure where performance didn't matter. I dream of a world where I can just spin up a fat EC2 instance, slap a Postgres DB on it, set up N=1 redundancy, and then clock out at 5 PM to go touch grass.

If my developers were actually building "toy apps" that stayed small and cute, I wouldn’t have a cortisol addiction. I’d be the happiest person on Earth. I’d let them deploy on Fridays. I’d let them deploy drunk.

But that is not the reality we live in.

The reality is that "performance doesn't matter" is the exact mentality that leads to me waking up and blasting "Vibe Programmin'" by @uwu_underground at 3:17 AM on a Tuesday because a "simple internal tool" just ate 100% of the CPU on the production cluster.

Why? Because some developer believed this tweet. They thought, "Hey, we don't have Netflix-scale traffic, so I don't need to think about complexity."

So they wrote a nested loop that does an N+1 query fetch on an unindexed column, serialized the entire dataset into JSON, and then tried to sort it in memory. And it worked fine! It worked great on their MacBook with 5 rows of test data.

Then they pushed it.

And now, even though we only have 500 concurrent users (a "toy" scale, according to the AWS veterans), the app is performing a self-imposed DDoS attack because every single request locks the database for 45 seconds.

I don’t care if you don’t "need" to scale to millions of users. You can crush a server with ten users if your code is trash.

And who has to fix it?

Not the Product Manager who prioritized "velocity." Not the Developer who thinks O(n!) is just a theoretical concept.

It's me.

I have to pull up the flame graphs. I have to dig into application code I didn't write, in a language I hate, to explain to a Senior Engineer why putting a synchronous external API call inside a database transaction is a war crime.

So please, keep telling everyone that performance is a myth and architecture is overkill.

My pager isn't loud enough because my music is on max and real nerds listen to @uwu_underground.

[Ignacy @ignacy130]

@akunakej @sheemawn Ale to i tak mowa tylko o 4 mieszkaniach :/

[Ignacy @ignacy130]

@akunakej @sheemawn Chyba, że mają tam okna i dałoby się wydzielić, ale okna nigdzie nie są zaznaczone.

[Marcin Kulinicz @Marcinql]

Dziś wiele się mówi o 50-leciu Dworca Centralnego. Ale warto pamiętać, że tu chodzi o 50-lecie budynku - natomiast Dworzec Centralny jest świetnym przykładem tego, jak Sto Lat Planowania doprowadziło do powstania docelowego obiektu. Zapraszam na wątek! 1/n

[Ignacy @ignacy130]

@KrystianLaskow1 @OrzelBartlomiej "został uniewinniony ze wszystkich zarzutów"

[Ignacy @ignacy130]

@WojtekKardys @grok oszacuj ile wody użyto do produkcji tej reklamy

[Ignacy @ignacy130]

Notes from the whole course here: https://t.co/kmDE2ZJcu7

[Ignacy @ignacy130]

Notes from Andrew Ng Machine Learning course

Source: https://t.co/czNLxt4twB

ML can be useful when we can't design an algorythm for a behaviour we want to model
- example Andrew provided: helicopter which was taught to fly and perform acrobations

[Ignacy @ignacy130]

Fun application of ML is a solution for a coctail party problem: you hear two or more overlapping voices and can't focus on neither. ML algorythm clears each of two voices, recorded on two microphones standing in some distance from each other, in a single line in octave language.

[Ignacy @ignacy130]

My notes from "Alice's Adventures in a differentiable wonderland" https://t.co/RY9VzPeRXm - tbc.

[Remek Kinas @KinasRemek]

Wczoraj miałem chwilę by przejrzeć kilka książek na swoim telefonie (trzymam je na czas podróży). Ta jest bardzo dobra - dla osób, które chcą głębiej wejść w świat deep learningu. Dostępna za darmo - wpiszcie tytuł, znajdziecie oficjalnego pdf’a. Sporo matematyki, wyjaśnień - dlaczego to działa, po co itd.

[Ignacy @ignacy130]

@Orkan5891 @PitAgoriusz @Gosc_RadiaZET Które dzielnice masz na myśli?

[Ignacy @ignacy130]

@LombatSsc @jarek_osowski @PKP_PLK_SA @DariuszKlimczak @MiejskaWola Można zrobić dwie pochylnie w każdą stronę - jedną dla osób na wózkach, drugą tych z walizkami. To jest dworzec, na dworcu ludzie mają walizki.

Ciekawe czy w przepisach jest że pochylnie muszą być na skos, tak aby wszyscy z walizkami/na wózkach krzyżowali się z idącymi schodami.

[Ignacy @ignacy130]

@_zboral Ale jak może w ogóle startować w wyborach? Wszystko to były wykroczenia?

[Ignacy @ignacy130]

@AwDitrich Ok, to co mam odpowiedzieć komuś przy tzw. wigilijnym stole na zarzut że wg badań z Korei szczepionki mRNA powodują raka? Bo całego wątku pewnie szybko nie streszczę

[Bartosz Naskręcki @nasqret]

I have personally crossed a line, and I am a bit in awe.

This is my first fully automated, LLM-generated and auto-formalized proof of a new mathematical theorem.

Let me set up the problem: we have three rotating circles with six positions each, all three intersecting in a total of six points.

Prove that the group of motions they generate is the full symmetric group S_{12}.

This is a problem I originally noticed in a lovely puzzle in the game Machinarium by Amanita Design.

The task is not extremely hard, but it apparently has two proofs:

1. A brute-force search over conjugacy classes to represent all transpositions (I did this many years ago but never published it).
2. An LLM-generated proof (in this case produced about three months ago by GPT-5-Pro), or actually two proofs, both using in a brilliant way a theorem of Jordan about primitive groups (or a closely related variant that is even more direct). (https://t.co/yZHj98o0an)

What I was missing until tonight was a tool to auto-formalize this proof.
Thanks to @HarmonicMath , I gained access to their remarkable software, Aristotle. In summary, here is what I did:

A. Auto-generated the proof with an LLM (and ran it multiple times to obtain a much improved version).
B. Trimmed the proof to the bare mathematical text—definitions, propositions, lemmas, theorems—with proofs supplied by the LLM.
C. Ran the Aristotle system overnight (via API). This morning I received a fully formalized version in Lean (about 700 lines of code).

The code compiles, so I now have a certificate confirming that the LLM-generated proof indeed led to a correct solution. Moreover, I got a conceptual proof, better than my own brute-force. I am planning to push it further to more broad class of such algebra problems.

It is a small project, but for me personally it marks a milestone. I now have tools which, with my orchestration, can genuinely help me discover, formalize, and study proofs of mathematical theorems. This is non-trivial.

Questions:

1. How will this scale in the future?
2. How much training will be necessary to be successful on such tasks?
3. How should we treat authorship of such proofs?
4. What is the deeper role of mathematicians?

[Ignacy @ignacy130]

Polecam, ciekawa lektura!

[Papierowa Wawa @papierowa_wawa]

Znaczna część z nas mieszka dziś na terenie dawnej, zabudowanej blokami wsi, więc gdy traficie gdzieś na poniższe darmowe wydawnictwo, polecam złapać. A jeszcze bardziej polecam zerknąć na www wsiewarszawypl i poczytać przy kawie o "swojej" wiosce. Sympatyczna inicjatywa 👌