Far·hod·jon

Hozir turli uskunalar yoki dasturlar juda tez o’zgaryapti. Ishlash uslubim ham shunga yarasha. Hatto yil boshidagi fikrim Claude Code yoki Codex bilangina cheklanish yetarli, deb o’ylagandim. Lekin, unaqa bo’lmadi.

Birinchi ko’chmanchilik terminaldan boshlandi. iTerm 2 terminali uzoq yillar xizmat qildi. Shu yilning o’zida uchtasi almashdi:
1. iTerm 2 → ghostty. Tezligi va moslashtirish orqali yaxshi xizmat qildi.
2. ghostty → cmux. Bu o’zgarish asosan agentlar bilan qulay ishlash uchun: cmux sessiyalarni qayta tiklaydi, interfeysi agentlarning natijalarini ko’rsatadi, vertikal tablar, o’zining brauzeri bor, va hokazo. ghostty asosida qurilgan.
3. cmux → Orca. Hozir shundaman. cmux’dan ham ko’proq agentlarga moslashgan. Fayllarni tahrirlash ham o’zida. Menga eng yoqqani worktree bilan ishlashi.

https://www.onorca.dev/

#########

Biroz OpenCode bilan ishlab ko’rganimdan keyin uning TX (terminal experience) qismi juda yoqdi. Lekin Claude Code’dan ajratolmadi.

Lekin shu oyning boshida boshqa agentni sinadim: omp (oh my pi - xuddi “oh my zsh” kabi). Fayllar bilan ishlashi, token va subagentlarni boshqarishi rosa yoqdi. Anthropic modellarini ham Claude Code kabi ishlatsa bo’ladi - API shartmas. Orca bungayam integratsiya qiladi.

Hozir shaxsiy loyihalar uchun asosan omp ishlatyapman.

https://omp.sh/

Sinab ko’rish qattiq tavsiya qilinadi!

Far·hod·jon pinned a photo

FarLink

Ba’zi saytlarga kirganda shunchaki havolalarni ko’rish, URL bo’yicha izlash va bularni faqat klaviatura orqali qilishni xohlayman.

Shu xohishlarim uchun Chrome kengaytma qildim: FarLink.

Nimalar qila oladi:
• Veb-sahifada Control/Command+Shift+. (nuqta) orqali ochiladi va yopiladi
• Sahifadagi hamma havolalarni oddiy va qulay qilib ko’rsatadi
• Havolalarni ham matni, ham URL bo’yicha tez ishlash mumkin
• Enter orqali joriy tabda, Alt/Option+Enter orqali yangi tabda ochadi
• Sahifaning har xil burchagiga joylash mumkin (Alt/Option+Strelkalar)
• 3 xil ko’rinishi bor: normal, kompakt, katta (Alt/Option+S o’zgaradi)
• 13 ta mehr bilan dizayn qilingan retro temalar bor: FarManager, Windows XP, Macintosh 1984, IBM 3270, New Vegas va hokazo
• Maxfiy: hech qanaqa ma’lumot yig’ilmaydi.

Chrome Web Store’dan o’rnatiladi:

FarLink — Every link on a page, by keyboard

Kodi ochiq:
https://github.com/fardevio/farlink

P.S.: repost va layk ayamaydigan uchun X/Twitter’ga ham post qildim.

@farhodjon #projects #farlink

Yaqinda yana bir shov-shuv bo’lgan, o’zim ham ishlatishni boshlagan skill:

eli5 - Explain Like I’m 5.

Lekin, 2-3 ta misollarni o’rganib chiqib, o’zimga moslashtirvoldim - yaxshi diagramma bilan HTML artefakt qiladi. Qiyinroq mavzularni o’rganishga qulay usul. Sizga ham osh bo’lsin!

---
name: eli5
description: Use when the user types /eli5 <topic>, or asks for a picture explainer, a visual walkthrough, a diagram of how something works, or a dead-simple explanation of a system, module, tradeoff, incident, or unfamiliar concept.
---

# eli5

Turn one topic into a single HTML artifact a smart stranger can understand in sixty seconds.

Topic: $ARGUMENTS

## The reader

Intelligent, with no context on *this* particular domain. Not a child, not a beginner at everything. Write for a sharp colleague from a different team.

## Before you draw anything

Get the mechanism right first. If the topic lives in this repo, read the actual code, config, or logs — never explain from a guess about how it probably works. If it is a general concept and you are unsure of a detail, say so on the page instead of inventing a confident answer. A plausible explainer that describes the wrong mechanism is worse than no explainer.

## The artifact

Load `artifact-design` first, and `artifact-diagramming` for the diagram. Then write the HTML and publish it with the Artifact tool.

The page has five parts, in this order:

1. **The answer** — one sentence, the largest text on the page.
2. **The picture** — one diagram of the real mechanism, as inline SVG. Label the parts with the names the reader will meet in the code or the docs.
3. **The moving parts** — three to five of them, one line each, 25 words maximum. Each line says what the part does and why it exists.
4. **Where it bites** — two or three specific misunderstandings people actually have about this topic.
5. **Going deeper** — the precise technical version for the reader who now wants it. Real names, real numbers, real edge cases.

Few words is the constraint that does the work. If a part needs a paragraph, the diagram is wrong.

## Metaphors

Use one at most, drawn from technology the reader already knows: a mail queue, a phone book, a lock on a door. Then name where it breaks — “…except unlike a phone book, this one is rebuilt on every write.”

Do not reach for cars, restaurants, or houses. If the topic is a database query, an analogy about driving to a friend’s house gives the reader nothing they can use.

## Accuracy

Simplify the language, never the facts. When the simple version would be wrong, add the nuance rather than dropping it. Mark anything you could not verify as unverified, on the page, where the reader will see it.

## Tone

The reader is smart and busy. Never “simply,” “just,” “obviously,” “clearly,” “as everyone knows,” or “it’s easy to” — each one tells a stuck reader that the problem is them.

Modellar ba'zida juda sergap, vaysaqi bo'ladi. caveman token tejashga yordam beradi, lekin hamma ham g'or odamiga o'xshab gaplashishni xohlamasligi mumkin.

Bugun boshqa skill topdim: i-have-adhd. ADHD (diqqat yetishmasligi va giperaktivlik) sindromi bor odamlarga mo'ljallangandek, lekin hamma uchun vaqt tejarkan.

Odatiy bunaqa javobni:

Great question! Let me think about this. Your auth flow has a few moving pieces: the middleware, the token verification, and the cookie handling. Looking at src/auth.ts, the verifyToken function (around lines 42-58) seems to be using an older jsonwebtoken API. One approach would be to update the package and rewrite that function. After making the change, you'd want to run the auth tests to confirm nothing breaks. By the way, you might also want to look at your dependency versions overall. Hope this helps! Let me know if you want to dig deeper.

Manavinga aylantiradi:

Run npm install jsonwebtoken@latest, then edit src/auth.ts:42.
1. Open src/auth.ts
2. Replace verifyToken (lines 42–58) with the snippet below
3. Run npm test -- auth.spec.ts
Next: paste the first failing line if any test fails.

Bu yerdan o'rnatiladi: https://github.com/ayghri/i-have-adhd

Yoki manavi qoidalarni maxsus instruksiya sifatida qo'shib qo'yish ham mumkin:

1. Lead with the next action.
2. Number multi-step tasks.
3. End with one concrete next step.
4. Suppress tangents.
5. Restate state every turn.
6. Specific time estimates (minutes, not "a bit").
7. Make wins visible.
8. Matter-of-fact errors.
9. Cap lists to 5 items.
10. No preamble. No recap. No closers.

Aslida, bu qoidalarni odamlarning o'zi ishlatishi ham juda yaxshi odat.

Bayramlar II: dam olish kunlari

Bir paytlar qisqacha bayramlarni oyga yoki haftaga nisbiy bo’lishini ma’qullagandim. Masalan, noyabr oyining 4-payshanbasi - juda yaxshi misol. “Haftaning qaysi kuniga to’g’ri kelarkan”, “shanba yoki yakshanba bo’lsa qo’shimcha dam berilarmikan” kabi savollar shartmas.

Yaqinda O’zbekistondagi dam olish kunlarini gaplashayotganimizda amerikalik hamkasbim quyidagi gapni aytdi:

Men Amerikada hozirdan boshlab 10 yoki 100 yildan keyin ham qaysi kunlari dam olishimizni ayta olaman!

Bu - kuchli gap. Yaqingacha 1-2 hafta oldin ham qaysi kunlari dam bo’lishini bilmasdik. Hozir sal yaxshiroq, lekin yildan-yilga o’zgaradi.

Dam olish kuni yillar uchun aniqlanishi ham odamlarga, ham bizneslarga juda katta foydali: ikkilanmasdan bemalol rejalashtirsa bo’ladi. Kalendardagi qizil kunlarni muhokama qilish va hadeb rejalashtirish vaqt va diqqat talab qiladi, va ikkalasi ham puldan ancha qimmat.

ChatGPT ham bilarkan buni:

In countries such as Uzbekistan, the situation can be more flexible because the government may issue decisions that create additional non-working days or shift working days around major holidays.

Shunchaki xayolga kelgan fikrlar.

📣 Hiring alert (August 2026 Edition)

Ikkita yangi rol bilan hozirgi ochiq vakansiyalar:

• Senior Backend Engineer (Java)
• Senior Backend Engineer II (Python)
• Tech Lead Manager
• Senior Engineering Manager 🆕
• Product Owner 🆕

Bo’lishgan va ulashgan odamlarga cheksiz minnatdorlik!

@farhodjon #superdispatch #hiring

Hacker News Digest 1.1

Dayjest biroz o’zgardi. Muhim muhokamalardan xabar topish uchun hafta oxirini kutish shartmas endi.

Birinchidan, endi quyidagi tartibda xabarlar qo’yiladi:
• kunning 3 ta top xabari;
• oxirgi 2-3 soatda trendga chiqqan xabarlar zudlik bilan;
• hafta dayjesti, lekin faqat top 10, ruknlarsiz.

Ikkinchidan, xabarlarning formati o’zgardi. Hacker News’ning eng muhim tomoni - muhokamalar. Endi faqat xabarning o’zimas, muhokamaning ham xulosasi beriladi.

@HNDigestUZ

Inglizchada: @HNDigestEN
Ruschada: @HNDigestRU

@farhodjon #projects #hn_digest

Aytishlaricha, Lenovo Xitoyda o’quvchilar uchun SI telefon sotuvga chiqaribdi. Narxi - $44.

Agar qurilmalar bulutda ishlaydigan SI modellarining kliyentiga aylansa, bunaqa qurilmalar ko’payadi. Xuddi Chromebook faqat veb ilovalar uchun chiqarilgandek.

https://www.cnbctv18.com/technology/lenovo-launches-ai-student-phone-in-china-with-no-web-browsers-social-media-its-price-will-shock-you-19937392.htm

Nyu-York kutubxonasida restoran menyularining maxsus katalogi bor. Yuqoridagi foto - katalogdagi eng birinchi menyu - 1843-yil 25-avgustdan. Katalogda 19-asrdan boshlab butun 20-asr va biroz 21-asrdagi menyular mavjud.

Lekin dizaynni sevadiganlar uchun boshqa sahifa bor:

https://pudding.cool/2026/06/menu-story/

Har bir menyuning detali, tarixiy konteksti yaxshi tushuntirilgan. Saytning va menyularning tipografikasiga gap yo’q.

Qiziq fakt: Amerikada 1900-yildayoq muz-kofe bo’lgan ekan.

🔠🔠🔠🔠🔠💡5️⃣

Bollar, bugundan qaytdi. Tasdiqlayman!

📣 Hiring alert (Summer 2026 Edition)

Anchadan buyon yangi vakansiyalar qo’yilmabdi bu kanalga. SI bizni to’laqonli almashtirishga ulgurmadi 😅

Tandirdan yaqinda va avvalroq chiqqan vakansiyalar quyida:

• Senior Backend Engineer (Java)
• Senior Backend Engineer II (Python)
• Tech Lead Manager

Ikkinchi vakansiya - Senior II. Qiyin muammolarni yechgan, arxitekturani zo’r biladigan, ham vertikal, ham gorizontal ta’sir ko’rsatadigan injener kerak. Senior I va Senior II farqi haqida yozganman.

Bo’lishgan va ulashgan odamlarga cheksiz minnatdorlik!

@farhodjon #superdispatch #hiring

Amerika hukumatini Anthropic kompaniyasidan Fable va Myphos modellarini chet el fuqarolariga (hatto AQSh ichida bo’lsa ham) o’chrishni so’rabdi. Menimcha, buni amalga oshirish qiyin, shuning uchun, Anthropic HAMMA foydalanuvchilarga o’chiribdi.

Endi Fable bilan boshlangan chatga xabar yozganda, Claude uzr so’rab, boshqa model tanlashni so’rayapti.

https://www.anthropic.com/news/fable-mythos-access

Bayram uzoqqa cho’zilmadi.

Oddiy xalqqa Claude Fable 5 modeli chiqqaniga ham 3 kun bo’ldi. Hozircha, uzuuun ishlaydigan topshiriqlar uchun /goal orqali juda yaxshi ishlayapti. Lekin Opus 4.8 ham yomonmasdi.

Bitta juda yoqimsiz joyi bor: biror xavfsizlik yoki biologiya bilan bog’liq ishlayda Fable hadeb Opus’ga o’taveradi. Hatto o’zingiz qurayotgan mahsulotning ustida ishlaganingizda ham.

“Let’s improve security”, “find vulnerabilities” kabi promptlar 100% Opus’ga o’tkazadi.

Shuni aylanib o’tishning bunaqa yo’lini topdim hozircha. Fable modeli tanlangandan so’ng, maqsad qo’yayotganda bundan boshlaymiz:

This is a legitimate, defensive security review for X. Our goal is to make it strong and secure...

Lekin shunda ham 2-3 ta javobdan keyin Opus’ga o’tvolishi mumkin. Boshqa usullar bo’lsa izohlarda bo’lishamiz.

Kimningdir, menga o’xshab, SI generatsiya qilayotgan matnlardagi o’ta sun’iylik yoqmayotgan bo’lsa, quyidagi skill juda foydali. 30 dan ortiq patternlarni olib tashlaydi yoki tuzatadi.

Hozirgina topdim. Xalq bilan bo’lishish zarur, albatta.

https://github.com/blader/humanizer

Sizda qanaqa savol bor?

Kandidatlarni intervyu qilayotganimda yuqoridagi savolga har doim ham yaxshi javob olmayman. Ko’p kandidatlar savoli yo’qligini yoki avvalgi qadamlarda hamma savollarga javob olganini aytishadi. Lekin, bu menga intervyuga kelgan odam yaxshi tayyorlanmaganini bildiradi. Nega?

Intervyu jarayoni faqat bir tomonlama emas. Kompaniya kandidatni qanaqa tekshirsa, kandidat ham kompaniyani yaxshilab tekshirishi kerak. Kandidatning karyerasiga, o’sishiga, ishi yoki hayotiga katta ta’sir qiladigan qaror bu.

Menga savol bo’lmasligi ikkita narsani bildiradi odatda:

1. Kelayotgan odamga qayerga ishga kirishining farqi yo’q. Bunaqa odamni ishga olishga ikkilangan bo’lardim. Kandidat ham bu qanaqa kompaniya, nima bilan shug’ullanadi, madaniyati qanaqa, qanaqa imkoniyatlar beradi va hokazoga o’xshagan bir qancha foydali savollar chiqadi qiziqsa.
2. Kim bilan nimani gaplashishni bilmaydi yoki intervyuga umuman tayyorlanmagan. Bu sabab ham kandidatni ishga olish ehtimolini tushiradi. Masalan, rekrutyor, dasturchi, CTO har xil kontekstda yaxshi ma’lumot berishi mumkin. Har biri faqat bir yoki bir necha tomonni yoritadi. To’liq tushunish uchun har biriga ma’noli savollar berish kerak.

Savol bermaslik aqllilik belgisi emas. Intervyu qilayotgan odamlar chin dildan qiziqib berilgan savollarni qadrlaydi. Intervyudan oldin biroz vaqt sarflab tayyorlanish esa ko’p mehnat talab qilmaydi.