Far·hod·jon

Koding agentlari bo’yicha so’rovnoma natijalari Yaqinda o’tkazilgan uchta so’rovnoma haqida fikrlarim bilan bo’lishsam. Yuqoriroqqa chiqib natijalarni ko’rishingiz ham mumkin. 1. Claude Code yutdi. Men Cursor’ni taxmin qilgandim, lekin u ikkinchi o’rinni oldi. Cursor va Antigravity juda yaqin bir-biriga. Antigravity bepulligini sabab sifatida taxmin qilishim to’g’rimi? 2. Koding agent uchun litsenziyani 72% odam o’zi to’larkan. 3. Agent ishlatmaydiganlar uchun eng asosiy sabab - narxi qimmatligi. Yuqorida ikkinchi punkt bu muammoni 100% yechishi kerak. 4. Agent ishlatmaydiganlar uchun “boshqa sabab” ikkinchi o’rinda, lekin uning tagida nima yotibdi, hayronman. Endi-chi? Birinchidan,dasturiy ta’minot ishlab chiqaradigan kompaniyalar dasturchilarga zudlik bilan koding agentlardan foydalanadigan muhit yaratishi kerak: litzensiya berish, o’rgatish, eng yaxshi tajribalarni olib kelish, kompaniya ichida entuziastlardan foydalanish. Cursor litzensiyasi oyiga $40 turadi. Oyiga $3000 oylik oladigan dasturchining mahsuldorligini 20% ga ko’paytirsa, kompaniya oyiga $560 qo’shimcha qiymat oladi ($3000×0.20-$40). Ikkinchidan, agentlarga juda skeptik qarayotgan yoki avvalroq sinab ko’rib hafsalasi pir bo’lganlarga 1-2 hafta davomida yaxshilab tajriba qilishni maslahat beraman. Yaxshiroq kontekst berib, to’g’ri yo’l ko’rsatilsa, hozirgi modellarning sifati ancha yaxshi. Ayniqsa, oldindan birga reja tuzib ishlaganda. Uchinchidan, asta-sekin eng kuchli dasturchilar ham turli agentlar bilan kod yozishni boshlayapti. Linus Torvalds Antigravity orqali Python’da kod yozyapti. Yozdagi intervyusida sun’iy intellekt yaxshi kod yozishiga ishonmagan DHH (Rails asoschisi) ham endi turli agentlardan foydalanyapti. Kuni kecha Redis asoschisi blog postida (hamma o’qishi kerak) quyidagi maslahatni beryapti: I have a single suggestion for you, my friend. Whatever you believe about what the Right Thing should be, you can't control it by refusing what is happening right now. Skipping AI is not going to help you or your career. Think about it. Test these new tools, with care, with weeks of work, not in a five minutes test where you can just reinforce your own beliefs. Find a way to multiply yourself, and if it does not work for you, try again every few months. Gemini tarjimasi: Sizga birgina maslahatim bor, do‘stim. Nima to‘g‘ri yoki noto‘g‘riligi haqidagi qarashlaringizdan qat’i nazar, hozirgi voqelikni shunchaki inkor etish orqali vaziyatni nazorat qila olmaysiz. Sun’iy intellektni chetlab o‘tish sizga ham, karyerangizga ham yordam bermaydi. Bu haqda o‘ylab ko‘ring. Yangi vositalarni shunchaki o‘z fikringizni tasdiqlash uchun besh daqiqa emas, balki haftalab vaqt sarflab, sinchkovlik bilan sinab ko‘ring. O‘z samaradorligingizni bir necha barobar oshirish yo‘lini toping va agar bu safar o‘xshamasa, har necha oyda qaytadan urinib ko‘ring. @farhodjon#ai#vibecoding

Claude Code Shu kunlarda Claude Code juda ommalashyapti. Ayniqsa, X/Twitter’da juda qiziq misollarni ko’rish mumkin: kimdir telefonda SSH orqali serverga kirib kod yozyapti, kimdir aqlli soati bilan gapirib kod yozyapti va hokazo. Hozir men uchta agent bilan ishlayapman: Claude Code, Codex (OpenAI) va Cursor. Lekin, X’dagi shov-shuvlarga qaraganda, Claude Code - eng yaxshi mahsulot (model sifatida unga savol yo’q). Agar Claude Code asoschisi - Boris Cherny’ning videosini ko’rmagan bo’lsangiz, zudlik bilan yarim soat vaqt sarflashni maslahat beraman. Ko’p narsalar umumiy qilib juda yaxshi tushuntirilgan. Mastering Claude Code in 30 minutes Video bergan yaxshi tajribalardan biri - butun tizim uchun ko’rsatmalar berib qo’yish. macOS’da ~/.claude/CLAUDE.md fayli tizimdagi hamma loyihalar uchun ishlaydi. Bu faylga umumiy qoidalar, shaxsiy stil, loyihadan agnostik bo’lgan ko’rsatmalarni yozib qo’yish mumkin. Quyidagi ko’rsatmalar Chris Dzombak blogidan olindi. Menimcha, juda yaxshi prompt. Shuni yaxshilab boraman. Sizga ham asqotadi, degan umiddaman. # Development Guidelines ## Philosophy ### Core Beliefs - **Incremental progress over big bangs** - Small changes that compile and pass tests - **Learning from existing code** - Study and plan before implementing - **Pragmatic over dogmatic** - Adapt to project reality - **Clear intent over clever code** - Be boring and obvious ### Simplicity - **Single responsibility** per function/class - **Avoid premature abstractions** - **No clever tricks** - choose the boring solution - If you need to explain it, it's too complex ## Technical Standards ### Architecture Principles - **Composition over inheritance** - Use dependency injection - **Interfaces over singletons** - Enable testing and flexibility - **Explicit over implicit** - Clear data flow and dependencies - **Test-driven when possible** - Never disable tests, fix them ### Error Handling - **Fail fast** with descriptive messages - **Include context** for debugging - **Handle errors** at appropriate level - **Never** silently swallow exceptions ## Project Integration ### Learn the Codebase - Find similar features/components - Identify common patterns and conventions - Use same libraries/utilities when possible - Follow existing test patterns ### Tooling - Use project's existing build system - Use project's existing test framework - Use project's formatter/linter settings - Don't introduce new tools without strong justification ### Code Style - Follow existing conventions in the project - Refer to linter configurations and .editorconfig, if present - Text files should always end with an empty line ## MCP Tool Use - Use Context7 to validate current documentation about software libraries - Use searxng if your primary Web Search or Fetch tools fail - Use Tavily ONLY when searxng doesn't give you enough information ## Important Reminders **NEVER**: - Use `--no-verify` to bypass commit hooks - Disable tests instead of fixing them - Commit code that doesn't compile - Make assumptions - verify with existing code **ALWAYS**: - Commit working code incrementally - Update plan documentation as you go - Learn from existing implementations - Stop after 3 failed attempts and reassess Loyihaning o’zi uchun Claude’ga bunaqa prompt berish yetarli: Scan the project and create a CLAUDE.md with build/test commands and code style guidelines based on what you find. Happy vibe coding! 🧑‍💻 @farhodjon#ai#vibecoding