Forwarded from Otabek’s I/O
Sizga sovg'am bor.
Bugun yakshanba, dam olish kuni. Algoritmlar bo'yicha bir paytlar o'tgan jonli darslarimni sizga o'rganishga bermoqchiman. Yana bunday darslarni o'tish uchun vaqt va sharoit topilishi bilan bularni yanada yaxshiroq qilishga harakat qilamiz.
1. Algoritmlarga kirish
2. Diskret Matematika
3. Array & Strings
4. Recursion
5. Stack & Queue
6. Graph
7. Tree
8. Deque
9. Linked List
10. Sorting
11. Hash Table
12. Searching
12. Sorting 2-qism
13. Dynamic Programming
14. Greedy Algorithms
Darslar yoqgan bo'lsa va fikringiz bo'lsa bemalol donat qilishingiz va fikringizni qoldirishingiz mumkin. Sizdan kelgan 5000 so'm va fikr ham men uchun juda ko'p support. Ziyo tarqating.
Bugun yakshanba, dam olish kuni. Algoritmlar bo'yicha bir paytlar o'tgan jonli darslarimni sizga o'rganishga bermoqchiman. Yana bunday darslarni o'tish uchun vaqt va sharoit topilishi bilan bularni yanada yaxshiroq qilishga harakat qilamiz.
1. Algoritmlarga kirish
2. Diskret Matematika
3. Array & Strings
4. Recursion
5. Stack & Queue
6. Graph
7. Tree
8. Deque
9. Linked List
10. Sorting
11. Hash Table
12. Searching
12. Sorting 2-qism
13. Dynamic Programming
14. Greedy Algorithms
Darslar yoqgan bo'lsa va fikringiz bo'lsa bemalol donat qilishingiz va fikringizni qoldirishingiz mumkin. Sizdan kelgan 5000 so'm va fikr ham men uchun juda ko'p support. Ziyo tarqating.
Anthropic AI engineer just showed how to give AI agents real memory in 4 steps - and it changes everything
in 28 minutes he shows exactly how agents can remember across sessions, completely free
worth more than any $500 AI engineering course
here's what he covers:
• why agents forget everything between sessions
• memory stores - agents read, write across sessions
• dreaming - agents that improve their own memory
• 95% cache hit rate, so it stays cheap
most people are still copy-pasting context into every new chat - while the people who figured this out are building agents that get smarter every single night
watch full video then read article below
in 28 minutes he shows exactly how agents can remember across sessions, completely free
worth more than any $500 AI engineering course
here's what he covers:
• why agents forget everything between sessions
• memory stores - agents read, write across sessions
• dreaming - agents that improve their own memory
• 95% cache hit rate, so it stays cheap
most people are still copy-pasting context into every new chat - while the people who figured this out are building agents that get smarter every single night
watch full video then read article below
Media is too big
VIEW IN TELEGRAM
Amazon AI engineer just showed how to take Claude apps from prototype to production on AWS.
19-minutes. Free. By the Amazon AWS team.
Claude Code → Bedrock → fully deployed, scalable, secure.
Worth more than any $500 vibe-coding AI course.
19-minutes. Free. By the Amazon AWS team.
Claude Code → Bedrock → fully deployed, scalable, secure.
Worth more than any $500 vibe-coding AI course.
Media is too big
VIEW IN TELEGRAM
Forwarded from JR TwitGram 🥂
Kurslare esa endi hamma uchun ochiq.
Building Your Own Web Framework in Python va Building Your Own ORM in Python
Building Your Own Web Framework in Python va Building Your Own ORM in Python
Forwarded from JavaHere's Blogs 🚀
Optimizatsiya 2 (72h -> 1h)
Pipeline bor. Bu pipeline har kuni bir kunlik ma'lumotlarni process qilib turishi kerak edi. Har safar 8 soat run bo'lardi.
U bir databasedan bir nechta table larni olib, birlashtirib, filterlab, boshqa bir databasedagi bir table ga yozadi.
Menga yangi bir column qo'shib uni kerakli ma'lumot bilan to'ldirishimni so'radi.
Shunday qildim, code ham review va approve bo'ldi.
Lekin kod pre-prod environmentga yetganda, naqd 72 soatdan ko'proq ishlab, time out bo'ldi.
Sababi, menga kerakli ma'lumotni query qilish uslubim xato bo'lgan ekan.
Qanday xato? Agar kerakli ma'lumot juda katta bo'lsa uni bir dona query orqali olib kelmasdan, alohida kichik pipeline bilan olib kelinadi.
Menga kerakli ma'lumot kichik (bir necha yuz megabayt) bo'lgani uchun uni batch emas streaming orqali, ya'ni, bitta queryda olib kelinadigan qildik.
Shunda 72h -> 25h bo'ldi.
Lekin bu ham yaxshi natija emasdi.
Hammani o'zini ishi bor, buniyam kimdir to'g'irlash kerak.
2 ta yo'limiz bor edi,
1. Kompyuterlarni kuchaytirish (cpu, ram)
2. Kodni optimizatsiya qilish
Mendan kodni optimizatsiya qilishimni so'radi.
Muammo - menga kerakli bir necha yuz megabaytli data bilan bir necha terabayatli datani join qilishda edi.
Shu qismini o'zi 17soat olayotgandi.
Men qilgan yechim:
Join qilish o'rniga, haligi kichik datani hashmapga soldim. Terabayt datani parallel yurib chiqib, har bir row uchun o'ziga mos datani hashmapdan olib to'ldirib chiqdim.
Qarasam huddi shunday ishni qilishim mumkin bo'lgan yana bir joy bor ekan. Uniyam shunday qildim.
Natija:
Endi pipeline ~1soat da ishni tugatyabdi, hamma xursand.
Olovni yondirgan menman. Lekin ayb menga yuklanmaydi, sababi jamoam bu kodni review qilib tasdiqlagan. Endi ayb hammaniki.
Optimizatsiyani men qildim, bu jamoaniki emas. Qiziq vaziyat.
Bu ishni qilish davomida Partitioning, Shuffling degan tushunchalar bilan tanishdim.
Balkim bular haqida alohida post yozarman.
Shunday qilib: 72h -> 1h.
Zo'r natija.
#tajriba
JavaHere
17.06.2026
Ishxona, Polsha
Pipeline bor. Bu pipeline har kuni bir kunlik ma'lumotlarni process qilib turishi kerak edi. Har safar 8 soat run bo'lardi.
U bir databasedan bir nechta table larni olib, birlashtirib, filterlab, boshqa bir databasedagi bir table ga yozadi.
Menga yangi bir column qo'shib uni kerakli ma'lumot bilan to'ldirishimni so'radi.
Shunday qildim, code ham review va approve bo'ldi.
Lekin kod pre-prod environmentga yetganda, naqd 72 soatdan ko'proq ishlab, time out bo'ldi.
Sababi, menga kerakli ma'lumotni query qilish uslubim xato bo'lgan ekan.
Qanday xato? Agar kerakli ma'lumot juda katta bo'lsa uni bir dona query orqali olib kelmasdan, alohida kichik pipeline bilan olib kelinadi.
Menga kerakli ma'lumot kichik (bir necha yuz megabayt) bo'lgani uchun uni batch emas streaming orqali, ya'ni, bitta queryda olib kelinadigan qildik.
Shunda 72h -> 25h bo'ldi.
Lekin bu ham yaxshi natija emasdi.
Hammani o'zini ishi bor, buniyam kimdir to'g'irlash kerak.
2 ta yo'limiz bor edi,
1. Kompyuterlarni kuchaytirish (cpu, ram)
2. Kodni optimizatsiya qilish
Mendan kodni optimizatsiya qilishimni so'radi.
Muammo - menga kerakli bir necha yuz megabaytli data bilan bir necha terabayatli datani join qilishda edi.
Shu qismini o'zi 17soat olayotgandi.
Men qilgan yechim:
Join qilish o'rniga, haligi kichik datani hashmapga soldim. Terabayt datani parallel yurib chiqib, har bir row uchun o'ziga mos datani hashmapdan olib to'ldirib chiqdim.
Qarasam huddi shunday ishni qilishim mumkin bo'lgan yana bir joy bor ekan. Uniyam shunday qildim.
Natija:
Endi pipeline ~1soat da ishni tugatyabdi, hamma xursand.
Olovni yondirgan menman. Lekin ayb menga yuklanmaydi, sababi jamoam bu kodni review qilib tasdiqlagan. Endi ayb hammaniki.
Optimizatsiyani men qildim, bu jamoaniki emas. Qiziq vaziyat.
Bu ishni qilish davomida Partitioning, Shuffling degan tushunchalar bilan tanishdim.
Balkim bular haqida alohida post yozarman.
Shunday qilib: 72h -> 1h.
Zo'r natija.
#tajriba
JavaHere
17.06.2026
Ishxona, Polsha
❤1
The most we want is what we cant have, and what we cant have is what we already had and lost
from the odyssey
from the odyssey
❤1