Michal Klos

2.2K posts

Michal Klos banner
Michal Klos

Michal Klos

@_spolsh

R&D, Pixels, Applied ML, Demoscene. ex-ByteDance/TikTok AI Lab. ex-games. Opinions are my own. https://t.co/H9UNZawZMI https://t.co/IRhP6FmIQG

Germany 가입일 Temmuz 2012
634 팔로잉604 팔로워
Michal Klos
Michal Klos@_spolsh·
@designworkplan @cory See Project PIWO PWr, Wroclaw, PL - clever control over lightbulbs in each building room and an annual show during University concert celebration
English
1
0
1
665
Cory Levy
Cory Levy@cory·
this is pretty cool some MIT students turned a building into a giant playable game of Tetris on Saturday at midnight rigged each window with LEDs MIT students are on a diff level
English
453
2.8K
27.3K
3.9M
Michal Klos
Michal Klos@_spolsh·
@valigo I know polish, easy. Now German vs enterprise java please.
English
0
0
4
94
Valentin Ignatev
Valentin Ignatev@valigo·
Let's play a game - win32 types vs Polish language: LPCWSTR PSZCZYNA WCSLEN WCZESNY LPCTSTR BYDGOSZCZ WSTRZAS HGDIOBJ DOWOD HWINSTA DLUGOSC LPCSTR DWORD KAL LPWSTR SZCZECIN PCWSTR BLAD PUHALF CHUJ UHALF
Polski
259
970
10.9K
469.3K
Sebastian Aaltonen
Sebastian Aaltonen@SebAaltonen·
@jmbynx Maybe because they last 10+ years without zero repairs. We sold ours when it was 10 years old to a friend. The car is now 14 years old. He had to fasten the exhaust once. Still no other issues. Can you drive a BMW for 14 years with no repairs?
English
11
1
128
12.3K
svadrb
svadrb@svadrb·
@_spolsh What does "Autoresearch" do? What's the main use?
English
1
0
0
16
Sebastian Aaltonen
Sebastian Aaltonen@SebAaltonen·
Threading our renderer with some Codex help. This is where LLM is very nice. It repeatedly starts your engine with simple test scene from command prompt (<10 iteration time) and iteratively debugs forward and validates/fixes code. Until it finds/fixes all issues.
Sebastian Aaltonen tweet media
English
9
0
70
6.4K
Michal Klos
Michal Klos@_spolsh·
@tanujDE3180 @SebAaltonen TBH this question baits Bloom filter answer to the point it kills creative problem solving skills Google would seek.
English
0
0
1
104
Tanuj
Tanuj@tanujDE3180·
@SebAaltonen Yes that is one of the solutions. But Google uses bloom filters for this usecase. It is a probabilistic data structure that do efficient lookup in O(1) time complexity.
English
4
0
50
56.7K
Tanuj
Tanuj@tanujDE3180·
Interviewer: You type a Gmail username and UI instantly shows "Username already taken" There are millions of users globally How does it check so fast?
Tanuj tweet media
English
299
195
6.8K
4.2M
Michal Klos
Michal Klos@_spolsh·
@SebAaltonen True and not true - more time and better product usually corelate, BUT 1 day ugly app marketed to correct community may win validation with 1 year app kept in the basement
English
0
0
0
62
Sebastian Aaltonen
Sebastian Aaltonen@SebAaltonen·
Unpopular opinion: Nobody will make money by single shotting a game or app. There's always somebody else who is willing to put in more effort. You spend 15 minutes, they spend a whole week. Of course their product is better. You beat their 1 week product with 1 month product, etc
English
45
26
557
31.7K
Michal Klos
Michal Klos@_spolsh·
I've shared k2py - 4k build system, an alternative to MS Visual Studio framework I've used for past 5 years github.com/sp0lsh/k2py Whats cool: 1 python file for all prod targets and zip into release 1 toggle for audio plugin: 4klang, oidos, sointu, none #demoscene #sizecoding
English
0
0
2
78
CobaltDragon
CobaltDragon@Cobalt__Dragon·
@FFmpeg Could you please also fragment the release builds into at least 20 executables in a nested interdependent directory structure of >= 657 files? Having a single exe that just does its job well is disconcerting. I'm much more comfortable when I don't know what's happening.
English
16
45
2.9K
132.3K
FFmpeg
FFmpeg@FFmpeg·
FFmpeg is moving to Rust 🦀 Our use of C and Assembly in FFmpeg has been an unacceptable violation of safety. FFmpeg will be running 10x slower - but we're doing it for your safety. All your videos will appear green - safety first, working software later.
English
1.6K
3.7K
44.5K
2M
Michal Klos
Michal Klos@_spolsh·
@Rahatcodes I though, a popup would tell the coder to take a walk and cool off :)
Michal Klos tweet media
English
0
0
2
294
rahat
rahat@Rahatcodes·
Very interesting follow up to this: There is a full mood classification for their insights but its employee only When an Anthropic employee gets frustrated, it pops up a prompt asking them to share their transcript, basically "hey you seem upset, wanna file a bug report?"
English
5
4
574
58K
rahat
rahat@Rahatcodes·
Claude Code has a regex that detects "wtf", "ffs", "piece of shit", "fuck you", "this sucks" etc. It doesn't change behavior...it just silently logs is_negative: true to analytics. Anthropic is tracking how often you rage at your AI Do with this information what you will
rahat tweet media
English
545
770
14.4K
1.4M
Stefan
Stefan@schteppe·
If a C api takes a buffer, but not buffer size… throw it into the trash
English
28
3
212
36.9K
Michal Klos
Michal Klos@_spolsh·
@SebAaltonen AI is good at rituals and scaffolding, it burped you a ritual piece over substance
English
0
0
2
135
Sebastian Aaltonen
Sebastian Aaltonen@SebAaltonen·
LLM explaining why it added yet another branch in perf critical inner loop: > "defensive fallback for callers that might render an unsynchronized scene" There's no such thing as "unsynchronized scene".
English
20
3
211
12.2K
Michal Klos
Michal Klos@_spolsh·
@olson_dan 'go mod vendor' is practical as simple deps cache when it takes too long to install on CI
English
0
0
0
30
Michal Klos
Michal Klos@_spolsh·
Vibe Coding feels more like BackstreetBoysing: 'Tell me why' x5 and then 'I want it that way'
English
0
0
0
50
Michal Klos
Michal Klos@_spolsh·
@wontfix @pshufb Also depends on monitoring i.e. Supabase has list of slowest queries and most frequent queries. Combined with above it can turn into quite methodological query perf cleanup
English
0
0
1
24
Andrew
Andrew@wontfix·
@pshufb Claude seems to know how to look at indexes and use EXPLAIN ANALYZE. Also query performance is very measurable, so the llms can just do the thing where they just try a bunch of different appproaches to see which one is empirically fastest.
English
1
0
3
117
Wojciech Muła
Wojciech Muła@pshufb·
Everybody is excited about "vibe coding", but how about database optimizations? Are there any good examples of using LLMs in optimizing DB structures, tuning queries, tuning DB options?
English
7
0
10
2.5K