Tuna Toksoz 𐰆𐰽𐰀

41.2K posts

Tuna Toksoz 𐰆𐰽𐰀 banner
Tuna Toksoz 𐰆𐰽𐰀

Tuna Toksoz 𐰆𐰽𐰀

@tehlike

Current: Meta Past: Google, MIT Aero/Astro, Bogazici I did robots, ads, robots, ads in that order. 🌓 🇺🇸

Mountain View, CA Katılım Ocak 2008
1.8K Takip Edilen2.2K Takipçiler
Paul Graham
Paul Graham@paulg·
If you have multiple kids you might be surprised by how different they look.
Paul Graham tweet media
English
119
11
1.8K
249.6K
fatih kadir akın
prompts.chat nasıl oldu da $6500 dolarlık fatura yaptı, hep vibecoding’ten işte ehehee falan diyenler gidip ilk günden beri tamamen açık kaynak olan kodu tamamen inceleyebilir, kaynak tüketen yer olduğunu bildiğiniz yer varsa gösterin. Boş yapmaya gerek yok, ben anlatayım: Hikaye şöyle: prompts.chat ilk versiyonu GitHub Pages’da statik idi. Dinamik hale getirince yeni versiyonu DigitalOcean App Platform’da yayınladım. Ben her işim için DO kullanıyorum sadece. Fakat yeni versiyonu duyurduğum gün DO app sürekli olarak çökmeye başladı. Hiçbir log yok vb. Destek de alamadım. Trafik de yoğun şekilde gelmeye başladığı için aynı günkü trafiği karşılaması için hızlıca Vercel’e geçirdim. Orada kaldı ve trafik yüksek olduğu için aylık 1200 dolar civarıydı geçmiş aylarda. Fakar sponsorluk ücretleri bu ücreti karşılıyordu. 2 ay önce kitapların PDF çevirilerini siteden yayınladım. CDN yerine function invocation yaptığım için o ay $3600 fatura ödedim. Sonra sitedeki tüm statikleri CDN’e çektim. Trafik yüksek olunca o statik PDF dosyalar can acıtıyor(muş). Site tamamen non-profit ve açık kaynak, sponsorluk harici geliri yok. Cepten yiyorum. Sponsorluklar da her an yok olabilecek şeyler olduğu için reklam koyarak bu giderleri bir şekilde giderebilir miyim diye alternatif yollar düşünüyordum, hosting’i değiştirmek de buna dahil. Fakat takdir edersiniz ki kolay olmuyor tek kişi olarak. Hep erteledim yani. Fakat bu ay ne olduysa bir anda /mcp endpointlerine 2 MİLYARA yakın request gelmiş, hepsi de edge request olarak ücretlendirilmiş. Bu da faturaya yansımış. Kısacası bu yüksek ücretin sebebi MCP istekleri. Bugün hemen MCP sunucusunu kapattım istekler devam etmesin diye. Sonrasında Rauch ulaşıp faturayı almayacaklarını ve projeyi OSS programına dahil edeceklerini söyledi. Olay örgüsü bu şekilde.
Türkçe
51
10
481
110.4K
Tuna Toksoz 𐰆𐰽𐰀
Use bloom filter, make it simpler lmao.
Abhishek Singh@0xlelouch_

Typed a Gmail username once and the UI instantly said: “Username already taken.” I asked an ex-Staff Google engineer the same problem (he was director of engineering in a startup i worked at), “You’re not doing an Elasticsearch query on every keypress, right?” He laughed. “No. That’d be a crime.” My classy approach: 1. Keep an in-memory trie of reserved usernames. 2. Update it async (delta pushes), not per keystroke. 3. UI checks locally in O(k) where k = username length. Numbers (why this is feasible): 1. Assume 2B usernames, avg length 10 chars. 2. Raw chars = 2B × 10 = 20B chars. 3. Even if you store 1 byte/char (not true in a trie, but baseline) that’s ~20GB just for characters. 4. A trie is about prefix sharing, so common prefixes collapse hard. Real memory is “nodes + edges”, not “strings”. 5. If we model ~1 node per char worst-case: ~20B nodes. - If a node is 8 bytes (tight packed arrays, bitsets, offset indices; no pointers), worst-case is 160GB. - With prefix sharing, you can easily cut multiples of that depending on distribution (gmail-like usernames are not random). 6. Shard by first 2 chars (36 possible: a-z, 0-9). 36² = 1296 shards. - Worst-case per shard: 160GB / 1296 ≈ 123MB. - Suddenly “instant check” fits in memory per front-end pod or edge POP. Yes, you can also do it with WebSockets: 1. Client streams “candidate username” events. 2. Server replies with availability. 3. Works fine, but now you’ve built a hot, stateful, low-latency service for… a UI hint. Most people will ship: 1. Elasticsearch prefix search. 2. Debounce 150ms. 3. Cache a bit. 4. Pray at peak signup traffic. And it works. But the trie approach is the kind of solution where the UI feels like magic tbh and it's something novel that i thought of. Things are just different at google scale.

English
0
0
0
44
fatih kadir akın
guys, i think i'm fucked up with prompts.chat. /cc @rauchg i think i need to exit this app and the repo before i bankrupt 😔 current sponsorships don't cover this.
fatih kadir akın tweet media
English
107
7
357
552.3K
Chayenne Zhao
Chayenne Zhao@GenAI_is_real·
the real answer nobody in this thread is giving: storing a photo costs instagram roughly $0.000005. at scale, storage is basically free. the expensive part isnt storing your selfies, its serving them - delivering the right photo to the right user in milliseconds, billions of times a day. thats the actual infrastructure cost. ads cover it 1000x over because meta built one of the most efficient serving stacks on the planet. the "you are the product" take is true but incomplete. the real story is that serving infrastructure at scale makes per-unit costs approach zero @xoaanya
Aanya@xoaanya

Interviewer: If apps like Instagram are free, how do they store billions of photos and videos without charging users anything?

English
8
30
1.1K
126.6K
Zach Tratar
Zach Tratar@zachtratar·
I can't wait to wear 5 different AI rings, 2 AI bracelets, an AI watch, and 2 AI necklaces. Why wear just an AI assistant when you could wear an entire AI council at once?
English
7
0
15
1.3K
Matt Turck
Matt Turck@mattturck·
Real life events: VCs to founders, during the board: "It's war time, guys. Time for founder mode" Founders: "Great. Then let's cancel the board dinner at [fancy restaurant] tonight and turn it into a work session for all of us in this conference room, with thai takeout" VCs: "nooooooo"
English
25
9
444
59.6K
Tuna Toksoz 𐰆𐰽𐰀
I might have vibe coded the most comprehensive (and one of the very few) key-value ORMs in existence using TiKV. And i don't even know how it works, just used my ORM experience from my NHibernate committer days. Now i should vibe code @RavenDB on this, right @ayende?
English
1
0
1
85
Tuna Toksoz 𐰆𐰽𐰀
That was one of the very few moments I felt very intense empathy - one that made me think my own kid was going through the same. Today I prayed for a lot of fathers and mothers. I will keep praying for them.
English
0
0
0
38
Tuna Toksoz 𐰆𐰽𐰀
My twitter timeline is people with a relative in a very hard health condition, and them asking for prayers. I prayed a lot today for them. Least i could do, i think. Today i almost cried when i remembered my friend's 1yo (at the time) going through lukemia.
English
1
0
0
52
Tuna Toksoz 𐰆𐰽𐰀
But yeah, it's a memorial for god's sake - people lost their lives for our country.
English
0
0
0
21