Chapterبەش 5 · Lectureوانە 5.1

Case Study: ArtCraft's AI-Built Open-Source Appsنموونەیەکی ڕاستەقینە: ئەپە کۆدکراوەکانی ArtCraft کە بە ژیریی دەستکرد دروستکراون

A case study of ArtCraft's free, open-source Craft Apps as of November 2026: what the 16 apps are and how finished they are, how one engineer built them with AI coding agents and which engineering rules still apply, Rust and WebAssembly, compiled apps versus Electron and web views, why compiled and local-first apps can be fast (with a fair Python-versus-C measurement), automation through CLI, JSON and MCP, permissive licences, clean-room design and trademarks, the 2026 fall in software shares and what still protects Adobe and Microsoft, and how to judge claims with dated evidence.

نموونەیەکی ڕاستەقینەی Craft Appsـی بێبەرانبەر و ArtCraft تا تشرینی دووەمی ⁦2026: 16⁩ چین و چەندە تەواو بوون، یەک ئەندازیار چۆن بە دروستی کردن و کام یاساکانی ئەندازیاری هێشتا کاریگەرن، Rust و ، کۆمپایلکراو بەرامبەر Electron و ، بۆچی کۆمپایلکراو و ناوخۆیی لە پێشدا دەتوانێت خێرا بێت (بە پێوانەیەکی دادپەروەرانەی Python بەرامبەر C)، لە ڕێگەی CLI، و MCP، ، و ، دابەزینی پشکی کۆمپانیاکانی لە ⁦2026⁩دا و ئەوەی هێشتا Adobe و Microsoft دەپارێزێت و چۆن بە بەڵگەی ڕێکەوتداریش بڕیار لەسەر بانگەشەکان بدەیت.

24 min read26 خولەک خوێندنەوە · 16 stepsهەنگاو

What you will learnچی فێر دەبیت

  • Describe what ArtCraft's Craft Apps are and how finished they are, using primary sources.
  • Explain how AI coding agents fit into a software process and which engineering work stays with people.
  • Explain Rust, WebAssembly, Electron and local-first design, and why compiled apps can be fast and light.
  • Read permissive licences, clean-room design and trademarks at concept level.
  • Judge claims about AI and the software market with dated evidence: true, false or not proven.
  • وەسفی بکەیت Craft Appsـی ArtCraft چین و چەندە تەواو بوون، بە بەکارهێنانی سەرچاوەی سەرەکی.
  • ڕوون بکەیتەوە چۆن دەچنە ناو پرۆسەی و کام کاری ئەندازیاری لای خەڵک دەمێنێتەوە.
  • Rust، ، Electron و دیزاینی ناوخۆیی لە پێشدا ڕوون بکەیتەوە و بۆچی کۆمپایلکراو دەتوانێت خێرا و سووک بێت.
  • ، و لە ئاستی بیرۆکەدا بخوێنیتەوە.
  • بە بەڵگەی ڕێکەوتداریش بڕیار لەسەر بانگەشەکانی و بازاڕی بدەیت: ڕاستە، هەڵەیە یان نەسەلمێنراوە.

Sixteen free apps, built with AIشازدە ئەپی بێبەرانبەر کە بە یارمەتیی ژیریی دەستکرد دروستکراون

In October 2026, news sites around the world wrote about ArtCraft, a small American company. Its Craft Apps are free programs for creative and office work: a photo editor, a video editor, a PDF tool, a word processor, a spreadsheet, slides, CAD and more. The reports said that one developer had built them with an AI coding model.

لە تشرینی یەکەمی ⁦2026⁩دا ماڵپەڕەکانی هەواڵ لە سەرتاسەری جیهان لەبارەی ArtCraft نووسییان، کە کۆمپانیایەکی بچووکی ئەمریکییە. ئەم کۆمپانیایە، کە ناویان Craft Appsـە، کۆمەڵێک بێبەرانبەرن بۆ کاری هونەری و ئۆفیس: ، دەستکاریکەری ڤیدیۆ، ئامرازێکی PDF، ، ، ، CAD و زیاتر. ڕاپۆرتەکان گوتیان یەک بە یارمەتیی مۆدێلێکی بۆ دروستی کردوون.

In this case study you check them as a software engineer would: what they really are, how they are made, why apps like them can be fast, and whether they really hurt Adobe and Microsoft. Facts are as of November 2026, and every number comes with its source and date.

لەم نموونە ڕاستەقینەیەدا وەک ئەندازیارێکی دەیانپشکنیت: بەڕاستی چین، چۆن دروستکراون، بۆچی وەک ئەمانە دەتوانێت خێرا بێت و ئایا بەڕاستی زیانیان بە Adobe و Microsoft گەیاندووە. زانیارییەکان هی تشرینی دووەمی ⁦2026⁩ن و هەر ژمارەیەک سەرچاوە و ڕێکەوتی خۆی لەگەڵدایە.

Six boxes group the 16 Craft Apps by kind of work. Pictures: PhotoCraft (photo editing), LightCraft (photo library, raw photos), VectorCraft (vector drawing). Video and motion: FilmCraft (video editing), EffectCraft (motion graphics, effects). Office and documents: WordCraft, GridCraft (spreadsheets), DeckCraft (slides), PdfCraft (PDF files). Design: DesignCraft (page layout), FigCraft (interface design). 3-D and engineering: TextureCraft, MaterialCraft, CADCraft. Sound and learning: SoundCraft and Craft Anything (see how software works). A filled dot marks the four early-alpha apps: PhotoCraft, PdfCraft, FigCraft and Craft Anything; an empty dot marks the twelve in development. Bottom: all 16 are free, open source, written in Rust, for Windows, macOS and Linux; 15 also run in a browser.شەش سندوق 16 Craft Appـەکە بەپێی جۆری کار کۆ دەکەنەوە. وێنە: PhotoCraft (دەستکاریی وێنە)، LightCraft (کۆکردنەوەی وێنە و وێنەی RAW)، VectorCraft (گرافیکی ڤێکتەری). ڤیدیۆ و جووڵە: FilmCraft (دەستکاریی ڤیدیۆ)، EffectCraft (گرافیکی جووڵاو و کاریگەری). ئۆفیس و بەڵگەنامە: WordCraft، GridCraft (بەرنامەی خشتەیی)، DeckCraft (سلاید)، PdfCraft (فایلی PDF). دیزاین: DesignCraft (ڕێکخستنی لاپەڕە)، FigCraft (دیزاینی نێوبەستە). سێ ڕەهەندی و ئەندازیاری: TextureCraft، MaterialCraft، CADCraft. دەنگ و فێربوون: SoundCraft و Craft Anything. خاڵی پڕ چوار ئەپی وەشانی ئەلفای سەرەتایی نیشان دەدات: PhotoCraft، PdfCraft، FigCraft و Craft Anything؛ خاڵی بەتاڵ ئەو دوازدە ئەپەی نیشان دەدات کە لە گەشەپێداندان. خوارەوە: هەر 16 بێبەرانبەر و کۆدکراوەن، بە Rust نووسراون، بۆ Windows و macOS و Linux و 15 دانەیان لە تۆڕگەڕیشدا کار دەکەن.
The 16 apps listed on getartcraft.com/apps in November 2026.ئەو ⁦16⁩ لە تشرینی دووەمی ⁦2026⁩دا لە ⁦getartcraft.com/apps⁩ ڕیز کراون.CêIT

How finished are they?چەندە تەواو بوون؟

Read the labels before the headlines. On ArtCraft's own page in November 2026, four apps are early alpha (PhotoCraft, PdfCraft, FigCraft and Craft Anything) and the other twelve are in development. PhotoCraft's README says it is not yet a Photoshop replacement for daily professional work; its biggest gaps are missing tools, professional workflows and plug-in compatibility. An independent test by the design magazine Creative Bloq also found the apps not ready to replace professional tools.

پێش سەردێڕی هەواڵەکان، نیشانەی سەر بخوێنەرەوە. لە ماڵپەڕی خودی ArtCraft لە تشرینی دووەمی ⁦2026⁩دا، چوار سەرەتایین (PhotoCraft، PdfCraft، FigCraft و Craft Anything) و دوازدەکەی تر لە گەشەپێداندان. READMEـی PhotoCraft دەڵێت هێشتا جێگرەوەی Photoshop نییە بۆ کاری پیشەیی ڕۆژانە؛ گەورەترین کەموکورتییەکانی ئامرازی نەبوو، ڕێگاکانی کاری پیشەیی و گونجان لەگەڵ . تاقیکردنەوەیەکی سەربەخۆی گۆڤاری دیزاینی Creative Bloqـیش بۆی دەرکەوت کە هێشتا ئامادە نین جێی ئامرازە پیشەییەکان بگرنەوە.

Their version numbers say the same. ArtCraft's launcher was 0.15.0, and PhotoCraft's October release was 0.8.2. In Semantic Versioning, a version that starts with 0 means early work that may change at any time; 1.0.0 is the first stable promise.

ژمارەی هەمان شت دەڵێت. Launcherـی ⁦ArtCraft 0.15.0⁩ بوو و تشرینی یەکەمی ⁦PhotoCraft 0.8.2⁩ بوو. لە وەشانبەندیی واتاییدا، کە بە 0 دەست پێبکات واتە کارێکی سەرەتایی کە هەر کاتێک دەکرێت بگۆڕێت؛ ⁦1.0.0⁩ یەکەم بەڵێنی جێگیرە.

A road with six numbered stations: idea, in development, alpha, beta, 1.0.0 stable, and mature: years of use. A box saying 12 Craft Apps, 'in development', points to station 2; a box saying 4 Craft Apps, 'early alpha', points to station 3; a box saying Photoshop and Word, improved for more than 30 years, points to station 6. A bracket under stations 1 to 4 says 0.y.z: anything may still change; a bracket under stations 5 and 6 says 1.0.0 and later: a stable promise. Bottom: Launcher 0.15.0, PhotoCraft 0.8.2 (October 2026), DeckCraft 0.15.0.ڕێگایەک بە شەش وێستگەی ژمارەدار: بیرۆکە، لە گەشەپێداندا، ئەلفا، بێتا، 1.0.0ـی جێگیر و پێگەیشتوو: ساڵانێک بەکارهێنان. سندوقێک کە دەڵێت 12 Craft App، «لە گەشەپێداندا»، ئاماژە بە وێستگەی 2 دەکات؛ سندوقێک کە دەڵێت 4 Craft App، «ئەلفای سەرەتایی»، ئاماژە بە وێستگەی 3 دەکات؛ سندوقێک کە دەڵێت Photoshop و Word زیاتر لە 30 ساڵ باشتر کراون، ئاماژە بە وێستگەی 6 دەکات. کەوانەیەک لە ژێر وێستگەکانی 1 تا 4 دەڵێت 0.y.z: هەر شتێک هێشتا دەکرێت بگۆڕێت؛ کەوانەیەک لە ژێر 5 و 6 دەڵێت 1.0.0 و دواتر: بەڵێنێکی جێگیر. خوارەوە: Launcher 0.15.0، PhotoCraft 0.8.2 (تشرینی یەکەمی 2026)، DeckCraft 0.15.0.
Status labels and version numbers tell you how finished software is.نیشانەی دۆخ و ژمارەی پیشانت دەدەن چەندە تەواو بووە.CêIT

How were they made? With AI coding agentsچۆن دروستکراون؟ بە ئەیجێنتەکانی کۆدنووسین

According to the reports, the apps were made by Brandon Thomas, a software engineer in Atlanta, with Anthropic's AI model Claude Opus 5.5, released on 22 September 2026. A French report says that eight AI agents worked at the same time. How much code the AI wrote and how much he wrote is not known. He has said he is not 'just some casual vibe coder', someone who accepts AI code without reading it.

بەپێی ڕاپۆرتەکان، Brandon Thomas، ئەندازیارێکی لە Atlanta، بە مۆدێلی Anthropic بە ناوی ⁦Claude Opus 5.5⁩ دروست کردوون، کە لە ⁦22⁩ی ئەیلوولی ⁦2026⁩دا بڵاو کرایەوە. ڕاپۆرتێکی فەرەنسی دەڵێت لە هەمان کاتدا هەشت کاریان کردووە. نازانرێت چەندی کۆدەکەی نووسیوە و خۆی چەندی. خۆی گوتوویەتی کە ئەو تەنیا «vibe coder»ێکی ئاسایی نییە، واتە کەسێک کە کۆدی بێ خوێندنەوە وەردەگرێت.

An AI coding agent works in a loop: it reads the task, changes the code, runs the tests, reads the errors and tries again (Agents and loops). The repositories show the human side too: instruction files for AI agents (AGENTS.md, CLAUDE.md) and thousands of commits; PhotoCraft alone had more than 11,000 in November 2026.

لە کار دەکات: ئەرکەکە دەخوێنێتەوە، کۆدەکە دەگۆڕێت، تاقیکردنەوەکان کار پێدەکات، هەڵەکان دەخوێنێتەوە و دووپاتە هەوڵ دەداتەوە (ئەیجێنتەکان و خولەکان). لایەنی مرۆییش پیشان دەدەن: فایلی ڕێنمایی بۆ (AGENTS.md، CLAUDE.md) و هەزاران ؛ تەنیا PhotoCraft لە تشرینی دووەمی ⁦2026⁩دا زیاتر لە ⁦11,000⁩ هەبوو.

An animation in six frames. Five boxes in a row: task and rules (AGENTS.md, an issue), AI agents (write and change code), tests and checks (cargo test, clippy), human review (the pull request), merge and release (main, then 0.x). Frame 1: a person writes the task and the rules for the agents. Frame 2: three numbered agents work at once. Frame 3: a test fails, a cross appears and an arrow sends the errors back to the agents. Frame 4: the agents fix the code and a tick appears under the tests. Frame 5: a person reviews the pull request and answers for it. Frame 6: the change is merged into main and released as early 0.x software, marked alpha.
An animation in six frames. Five boxes in a row: task and rules (AGENTS.md, an issue), AI agents (write and change code), tests and checks (cargo test, clippy), human review (the pull request), merge and release (main, then 0.x). Frame 1: a person writes the task and the rules for the agents. Frame 2: three numbered agents work at once. Frame 3: a test fails, a cross appears and an arrow sends the errors back to the agents. Frame 4: the agents fix the code and a tick appears under the tests. Frame 5: a person reviews the pull request and answers for it. Frame 6: the change is merged into main and released as early 0.x software, marked alpha.
The AI writes and fixes code; tests and people decide what goes in. کۆد دەنووسێت و چاکی دەکات؛ تاقیکردنەوەکان و خەڵک بڕیار دەدەن چی بچێتە ناوەوە.CêIT

What the engineer must still doئەوەی ئەندازیار هێشتا دەبێت بیکات

PhotoCraft's AGENTS.md (November 2026) is a list of rules for every contributor, human or AI. Almost every rule is a lesson of this course: tests for every change, parts that may depend only on lower layers, main changed only through reviewed pull requests, and only licences from an allow-list.

AGENTS.mdـی PhotoCraft (تشرینی دووەمی ⁦2026⁩) لیستێکی یاسایە بۆ هەموو بەشداربوویەک، مرۆڤ بێت یان . نزیکەی هەموو یاسایەکی وانەیەکی ئەم کۆرسەیە: تاقیکردنەوە بۆ هەر گۆڕانکارییەک، بەش کە تەنیا دەتوانن پشت بە خوارەوە ببەستن، گۆڕینی main تەنیا بە کە پێیاندا چوونەتەوە و تەنیا کە لە .

Two columns joined by arrows. Left: six rules from PhotoCraft's AGENTS.md, November 2026. Right: where this course teaches them. Every change needs tests, run them before you finish: automated testing. A part may depend only on lower layers: layers and business logic. main changes only through pull requests: GitHub and teamwork. The person who runs an agent answers for its work: AI assistants, lesson 1. Only licences on an allow-list, such as MIT and Apache-2.0: licences, this lesson. Study other editors only for how they behave and never copy their code: clean-room design, this lesson. Bottom: AI changes who types the code, not what makes it trustworthy.دوو ستوون کە بە تیر بەیەکەوە بەستراون. چەپ: شەش یاسا لە AGENTS.mdـی PhotoCraft، تشرینی دووەمی 2026. ڕاست: ئەم کۆرسە لە کوێ فێریان دەکات. هەر گۆڕانکارییەک تاقیکردنەوەی دەوێت: تاقیکردنەوەی خۆکار. هەر بەشێک تەنیا دەتوانێت پشت بە چینەکانی خوارەوە ببەستێت: چینەکان و لۆجیکی کاروبار. main تەنیا بە داواکاریی بەیەککردن دەگۆڕێت: GitHub و کاری تیمی. ئەو کەسەی ئەیجێنت کار پێدەکات بەرپرسیارە لە کارەکەی: یاریدەدەرەکانی ژیریی دەستکرد، وانەی 1. تەنیا مۆڵەتەکانی بەکارهێنانی ڕێگەپێدراو وەک MIT و Apache-2.0: مۆڵەتەکانی بەکارهێنان، ئەم وانەیە. لە ئامرازەکانی تر تەنیا هەڵسوکەوتیان ببینە و هەرگیز کۆدیان لەبەر مەگرەوە: دیزاینی ژووری پاک، ئەم وانەیە. خوارەوە: ژیریی دەستکرد ئەوە دەگۆڕێت کێ کۆدەکە دەنووسێت، نەک ئەوەی چی جێی متمانەی دەکات.
The project's rules for AI agents are ordinary software-engineering rules.یاساکانی پڕۆژەکە بۆ یاسای ئاسایی .CêIT

The file also says that the person who runs an agent is accountable for its pull requests. The tools changed who types the code, not what makes it trustworthy: clear requirements, automated tests, code review, and careful versions and releases.

فایلەکە هەروەها دەڵێت ئەو کەسەی کار پێدەکات بەرپرسیارە لە . ئامرازەکان ئەوەیان گۆڕیوە کێ کۆدەکە دەنووسێت، نەک چی جێی متمانەی دەکات: ڕوون، تاقیکردنەوەی خۆکار، پێداچوونەوەی کۆد و وەشان و بڵاوکردنەوەی وردبینانە.

Rust and WebAssembly: one code base, two kinds of programRust و WebAssembly: یەک کۆی کۆد، دوو جۆر بەرنامە

Every Craft App is written in Rust. A Rust compiler turns the source code into machine code before the program runs, so nothing has to be translated while you work. Rust has no garbage collector: the compiler checks ownership rules, so memory is freed at known points (how languages free memory). The Rust project describes the language as fast and memory-efficient, with no runtime or garbage collector.

هەموو Craft Appـێک بە Rust نووسراوە. کۆمپایلەری Rust پێش کارپێکردنی دەکات بە ، بۆیە لە کاتی کارکردنتدا هیچ شتێک پێویستی بە وەرگێڕان نییە. Rust بەکار ناهێنێت: یاساکانی خاوەندارێتی دەپشکنێت، بۆیە لە خاڵێکی زانراودا ئازاد دەکرێتەوە (زمانەکان چۆن بیرگە ئازاد دەکەنەوە). پڕۆژەی Rust ئەم زمانە بە خێرا و کەمخەرج لە وەسف دەکات، بێ runtime و بێ .

The same Rust code can also be compiled to WebAssembly (Wasm), a compact binary format that browsers run at near-native speed in a safe sandbox, next to JavaScript. That is how 15 of the 16 apps also open in a browser tab; FilmCraft, the video editor, is desktop only.

هەمان کۆدی Rust دەتوانرێت بۆ ⁦(Wasm)⁩ کۆمپایل بکرێت، شێوازێکی دووانیی چڕ کە بە خێرایییەکی نزیک لە سەر ئامێر لە و لە تەنیشت کاری پێدەکەن. بەم شێوەیە ⁦15⁩ لە ⁦16⁩ لە تابێکی دەکرێنەوە؛ FilmCraft، کە دەستکاریکەری ڤیدیۆیە، تەنیا بۆ سەر مێزە.

Rust source code (.rs) goes into the compiler (cargo build). One path leads to machine code for each system: Windows, macOS and Linux; it runs on the CPU and draws with the GPU through Metal, Vulkan or DirectX 12. The other path leads to WebAssembly (.wasm), which runs in a browser tab: a safe sandbox next to JavaScript, drawing with WebGPU. Bottom: same code, two outputs, a desktop program and a web version; 15 of the 16 Craft Apps take both paths, and FilmCraft is desktop only.کۆدی سەرچاوەی Rust (.rs) دەچێتە ناو کۆمپایلەر (cargo build). ڕێگایەک دەچێت بۆ کۆدی ئامێر بۆ هەر سیستەمێک: Windows، macOS و Linux؛ لەسەر CPU کار دەکات و بە GPU لە ڕێگەی Metal، Vulkan یان DirectX 12 وێنە دەکێشێت. ڕێگاکەی تر دەچێت بۆ WebAssembly (.wasm)، کە لە تابێکی تۆڕگەڕدا کار دەکات: ژینگەیەکی داخراوی سەلامەت لە تەنیشت JavaScript، بە WebGPU وێنە دەکێشێت. خوارەوە: هەمان کۆد، دوو بەرهەم، بەرنامەیەکی سەر مێز و وەشانێکی وێب؛ 15 لە 16 Craft App هەردوو ڕێگا دەگرن و FilmCraft تەنیا بۆ سەر مێزە.
One code base can be compiled for the desktop and for the browser.یەک کۆی کۆد دەتوانرێت بۆ سەر مێز و بۆ کۆمپایل بکرێت.CêIT

Native, not wrapped: what is Electron?کۆمپایلکراو، نەک پێچراو: Electron چییە؟

Many desktop apps today are web pages in a frame. Electron builds desktop apps from HTML, CSS and JavaScript by putting Chromium (the open-source browser behind Chrome) and Node.js inside every app. Like a browser, an Electron app runs several processes: one main process and a renderer process for each window. A lighter way uses the system's own , like hybrid mobile apps.

ئەمڕۆ زۆرێک لە سەر مێز لاپەڕەی وێبن لە ناو چوارچێوەیەکدا. Electron سەر مێز لە ، و دروست دەکات، بەوەی Chromium ( پشت Chrome) و ⁦Node.js⁩ دەخاتە ناو هەموو . وەک ، Electron چەند کار پێدەکات: سەرەکی و وێناکردن بۆ هەر پەنجەرەیەک. ڕێگایەکی سووکتر خودی سیستەمەکە بەکار دەهێنێت، وەک ئەپە تێکەڵەکانی مۆبایل.

ArtCraft says its apps use neither Electron nor web views. Each one is a compiled program that draws its own windows on the graphics card; PhotoCraft, for example, draws with wgpu, a Rust library that talks to Metal, Vulkan, DirectX 12 or WebGPU.

ArtCraft دەڵێت نە Electron بەکار دەهێنن و نە . هەر یەکێکیان کۆمپایلکراوە کە پەنجەرەکانی خۆی بە دەکێشێت؛ بۆ نموونە PhotoCraft بە wgpu دەکێشێت، Rust کە لەگەڵ Metal، Vulkan، ⁦DirectX 12⁩ یان WebGPU قسە دەکات.

Everyday analogyنموونەیەک لە ژیانی ڕۆژانەوە

An Electron app is like a guest who brings his own samovar, cups and gas bottle to every visit to make tea. A compiled app uses the host's kitchen. Ten guests bring ten samovars.

Electron وەک میوانێک وایە کە بۆ هەر سەردانێک سەماوەر، پەرداخ و بوتڵی گازی خۆی دەهێنێت بۆ ئەوەی چا دروست بکات. کۆمپایلکراو چێشتخانەی خاوەنماڵەکە بەکار دەهێنێت. دە میوان دە سەماوەر دەهێنن.

Where the analogy stops workingسنووری ئەم نموونەیە

The guest's own samovar makes the same tea in every house: an Electron app looks and works the same on every system, and web developers can reuse their skills.

سەماوەرەکەی میوان لە هەموو ماڵێکدا هەمان چا دروست دەکات: Electron لەسەر هەموو سیستەمێک یەک شێوە و یەک کاری هەیە و وێب دەتوانن شارەزاییەکانیان دووپاتە بەکار بهێننەوە.

Three stacks of layers. Compiled app, for example the Craft Apps: app code as machine code, then draws its own windows on the GPU, then the operating system, then hardware (CPU, GPU, memory). Web-view app, a web page in a frame: app code in HTML, CSS and JavaScript, then the system's web view (shared), then the operating system and hardware. Electron app, for example many chat and editor apps: app code in HTML, CSS and JavaScript, then Node.js, then the Chromium browser engine, a copy inside every app, then the operating system and hardware. Bottom: Electron puts a whole browser engine and Node.js inside every app; a compiled app needs neither.سێ ستوون، هەر یەکەیان چەند چینێک لەسەر یەک. ئەپی کۆمپایلکراو، وەک Craft Apps: کۆدی ئەپ وەک کۆدی ئامێر، پاشان پەنجەرەکانی خۆی بە GPU دەکێشێت، پاشان سیستەمی وەگەڕخەر، پاشان ڕەقەواڵە (CPU، GPU و بیرگە). ئەپی پیشاندەری وێب، لاپەڕەیەکی وێب لە چوارچێوەیەکدا: کۆدی ئەپ بە HTML، CSS و JavaScript، پاشان پیشاندەری وێبی سیستەم (هاوبەش)، پاشان سیستەمی وەگەڕخەر و ڕەقەواڵە. ئەپی Electron، وەک زۆر ئەپی چات و دەستکاریکردن: کۆدی ئەپ بە HTML، CSS و JavaScript، پاشان Node.js، پاشان بزوێنەری وێناکردنی Chromium کە کۆپییەکی لە ناو هەموو ئەپێکدایە، پاشان سیستەمی وەگەڕخەر و ڕەقەواڵە. خوارەوە: Electron بزوێنەری وێناکردنێکی تەواو و Node.js دەخاتە ناو هەموو ئەپێکەوە؛ ئەپی کۆمپایلکراو پێویستی بە هیچیان نییە.
Three ways to build a desktop app, from the hardware up.سێ ڕێگا بۆ دروستکردنی سەر مێز، لە بەرەو سەرەوە.CêIT

Why compiled apps can be fast and lightبۆچی ئەپی کۆمپایلکراو دەتوانێت خێرا و سووک بێت

There are good reasons why a compiled app like the Craft Apps can start quickly and use little memory:

هۆکاری باش هەن بۆ ئەوەی کۆمپایلکراوی وەک Craft Apps بتوانێت زوو بکرێتەوە و کەم بەکاربهێنێت:

  • No translation while it runs. Machine code runs straight on the CPU. JavaScript must first be compiled by the browser while the app runs (just-in-time compiling).
  • Less to start. No browser engine and no Node.js have to load before the first window appears.
  • Less memory. There is no browser copy inside every app, and no garbage collection runs: Rust frees memory as soon as it is no longer needed.
  • Local files and the GPU. Opening a file does not wait for the internet, and the graphics card does the drawing.
  • لە کاتی کارکردندا وەرگێڕان نییە. ڕاستەوخۆ لەسەر کار دەکات. دەبێت سەرەتا لە کاتی کارکردنی لەلایەن کۆمپایل بکرێت (کۆمپایلکردنی کاتی جێبەجێکردن).
  • شتی کەمتر بۆ هەڵکردن. پێش دەرکەوتنی یەکەم پەنجەرە هیچ بزوێنەرێکی و ⁦Node.js⁩ بار ناکرێن.
  • کەمتر. هیچ کۆپییەکی لە ناو هەر نییە و کار ناکات: Rust هەر کە پێویست نەما ئازاد دەکاتەوە.
  • ناوخۆیی و . کردنەوەی چاوەڕێی ناکات و وێنەکێشانەکە دەکات.
Two rows of steps from opening an app to the first window. Compiled app: load the program, draw the window, ready. Electron app: load the program, start the browser engine and Node.js, load HTML, CSS and JavaScript, run the JavaScript, draw, ready. A note says: a sketch of the steps, not a measurement; fewer steps can mean a faster start and less memory, and real speed also depends on the code, the computer and the work.دوو ڕیز هەنگاو لە کردنەوەی ئەپێکەوە تا یەکەم پەنجەرە. ئەپی کۆمپایلکراو: بارکردنی بەرنامەکە، کێشانی پەنجەرەکە، ئامادە. ئەپی Electron: بارکردنی بەرنامەکە، هەڵکردنی بزوێنەری تۆڕگەڕ و Node.js، بارکردنی HTML، CSS و JavaScript، کارپێکردنی JavaScript، کێشان، ئامادە. تێبینییەک دەڵێت: وێنەیەکی هەنگاوەکانە، نەک پێوانە؛ هەنگاوی کەمتر دەکرێت واتای دەستپێکی خێراتر و بیرگەی کەمتر بێت و خێرایی ڕاستەقینە پشت بە کۆد، کۆمپیوتەر و کارەکەش دەبەستێت.
Fewer steps before the first window: a reason, not a measurement.هەنگاوی کەمتر پێش یەکەم پەنجەرە: هۆکارێکە، نەک پێوانە.CêIT

Can, not will. ArtCraft publishes no speed measurements (), so this lesson makes no claim about how fast its apps are. A slow algorithm is slow in every language (speed depends on the program), and a well-built Electron app can feel quick. Measure before you believe.

دەتوانن، نەک دەبن. ArtCraft هیچ پێوانەیەکی خێرایی () بڵاو ناکاتەوە، بۆیە ئەم وانەیە هیچ بانگەشەیەک لەبارەی خێرایی ناکات. خاو لە هەموو زمانێکدا خاوە (خێرایی پشت بە بەرنامەکە دەبەستێت) و Electron کە باش دروستکرابێت دەتوانێت خێرا هەست پێ بکرێت. پێش باوەڕکردن، بیپێوە.

A fair test you can run yourselfتاقیکردنەوەیەکی دادپەروەرانە کە خۆت دەتوانیت بیکەیت

Rust is not installed on our test computer, so we use C, which is also compiled ahead of time to machine code. Both programs count the prime numbers below 1,000,000 with exactly the same simple method: try every divisor up to the square root.

Rust لەسەر کۆمپیوتەری تاقیکردنەوەکەمان دانەمەزراوە، بۆیە C بەکار دەهێنین، کە ئەویش بە دەکرێت بە . هەردوو ژمارە سەرەتاییەکانی خوار ⁦1,000,000⁩ بە هەمان ڕێگای سادە دەژمێرن: هەموو دابەشکەرێک تا ژمارەکە تاقی دەکەنەوە.

primes.pyPYTHONType in a terminalلە تێرمیناڵدا بینووسە
# primes.py - count the prime numbers below a limit
def count_primes(limit):
    count = 0
    for n in range(2, limit):
        d = 2
        while d * d <= n:
            if n % d == 0:
                break
            d += 1
        else:
            count += 1
    return count

print(count_primes(1_000_000))
primes.cCType in a terminalلە تێرمیناڵدا بینووسە
/* primes.c - the same work, in C */
#include <stdio.h>

static int count_primes(int limit) {
    int count = 0;
    for (int n = 2; n < limit; n++) {
        int prime = 1;
        for (int d = 2; d * d <= n; d++) {
            if (n % d == 0) { prime = 0; break; }
        }
        count += prime;
    }
    return count;
}

int main(void) {
    printf("%d\n", count_primes(1000000));
    return 0;
}
Timing both programs (macOS)کاتی هەردوو بەرنامەکە ⁦(macOS)⁩BASHType in a terminalلە تێرمیناڵدا بینووسە
clang -O2 primes.c -o primes      # compile once, ahead of time
./primes                           # first run (the system checks a new program)
/usr/bin/time -p python3.13 primes.py
/usr/bin/time -p ./primes
/usr/bin/time -l python3.13 primes.py 2>&1 | grep "maximum resident"
/usr/bin/time -l ./primes 2>&1 | grep "maximum resident"
Explain the codeڕوونکردنەوەی کۆدەکە
Lineدێڕی 3-4

real is the time on the clock in seconds; user is CPU time spent in the program.

real کاتی کاتژمێرە بە چرکە؛ user ئەو کاتەیە کە لە بەسەری بردووە.

Lineدێڕی 5-6

The most memory each program used, in bytes: about 13.3 MB for Python and 1.4 MB for C.

زۆرترین کە هەر بەکاری هێنا، بە بایت: نزیکەی ⁦13.3 MB⁩ بۆ Python و ⁦1.4 MB⁩ بۆ C.

On our Apple M1 laptop the compiled program was roughly 70 to 100 times faster over several runs (Python took 6.2 to 8.9 seconds, C 0.08 to 0.09 seconds) and used about a tenth of the memory. Your numbers will differ. A fair test uses the same algorithm and the same computer, runs each program several times, and ignores the first run of a new program.

لەسەر ⁦(Apple M1)⁩ لە چەند جارێکدا نزیکەی ⁦70⁩ تا ⁦100⁩ هێندە خێراتر بوو (Python لە ⁦6.2⁩ تا ⁦8.9⁩ چرکەی خایاند و C لە ⁦0.08⁩ تا ⁦0.09⁩ چرکە) و نزیکەی یەک لە دەی Pythonـی بەکارهێنا. ژمارەکانی تۆ جیاواز دەبن. تاقیکردنەوەی دادپەروەرانە هەمان و هەمان کۆمپیوتەر بەکار دەهێنێت، هەر چەند جار کار پێدەکات و یەکەم جاری نوێ پشتگوێ دەخات.

Two bar charts from our test. Time to count the primes, best run: Python 6.21 seconds, C 0.08 seconds. Most memory used: Python 13.3 MB, C 1.4 MB. Below: primes below 1,000,000 (answer 78,498), trial division, several runs each; Apple M1, macOS 26.5, Python 3.13.17, Apple clang 21 with -O2, 10 November 2026. C was about 70 to 100 times faster here; your numbers will differ.دوو هێڵکاریی ستوونی لە تاقیکردنەوەکەمان. کاتی ژماردنی ژمارە سەرەتاییەکان، باشترین جار: Python 6.21 چرکە، C 0.08 چرکە. زۆرترین بیرگە کە بەکارهات: Python 13.3 MB، C 1.4 MB. خوارەوە: ژمارە سەرەتاییەکانی خوار 1,000,000 (وەڵام 78,498)، دابەشکردنی تاقیکاری، چەند جار بۆ هەر یەکێک؛ Apple M1، macOS 26.5، Python 3.13.17، Apple clang 21 بە -O2، 10ی تشرینی دووەمی 2026. لێرەدا C نزیکەی 70 تا 100 هێندە خێراتر بوو؛ ژمارەکانی تۆ جیاواز دەبن.
One measured example, with the machine and the date written next to it.نموونەیەکی پێوراو، کە ئامێر و ڕێکەوتەکەی لە تەنیشتی نووسراون.CêIT

Your files, your machine: local-first softwareفایلەکانی خۆت، ئامێری خۆت: نەرمەواڵەی ناوخۆیی لە پێشدا

ArtCraft's page says that everything runs locally on your own files. This idea has a name: local-first software (Martin Kleppmann and colleagues at Ink & Switch, 2019). The main copy of your work lives on your own device; a server, if there is one, only helps to sync and back up.

ماڵپەڕی ArtCraft دەڵێت هەموو شتێک لەسەر ئامێرەکەت و لەسەر خۆت کار دەکات. ئەم بیرۆکەیە ناوێکی هەیە: (Martin Kleppmann و هاوکارەکانی لە ⁦Ink & Switch⁩، ⁦2019⁩). کۆپیی سەرەکیی کارەکەت لەسەر ئامێری خۆت دەژی؛ ، ئەگەر هەبێت، تەنیا یارمەتی و دەدات.

  • Good: it works offline, files open at once, your work stays private, and your files can outlive the company.
  • Harder: you must make your own backups, editing one file together with others is harder, and you install updates yourself.
  • باش: بێ کار دەکات، دەستبەجێ دەکرێنەوە، کارەکەت تایبەت دەمێنێتەوە و دەتوانن لە کۆمپانیاکە زیاتر بژین.
  • قورستر: دەبێت خۆت دروست بکەیت، دەستکاریکردنی یەک لەگەڵ کەسانی تر پێکەوە قورسترە و خۆت دادەمەزرێنیت.

Cloud apps make the opposite trade (benefits and risks of the cloud).

لە هەوردا بە پێچەوانەوەیە (سوود و مەترسییەکانی هەور).

Two panels. Cloud-first app: your laptop shows only a view and talks over the internet to a company server that holds the main copy; with no internet you may wait or not open your file at all. Local-first app: your laptop holds the main copy; a dashed arrow leads to an optional backup and sync; it works offline and opens at once, and you make the backups. Bottom: the seven ideals of local-first software (Ink & Switch, 2019): fast, multi-device, offline, collaboration, longevity, privacy, user control.دوو بەش. ئەپی هەور لە پێشدا: لاپتۆپەکەت تەنیا پیشاندانێکە و لە ڕێگەی ئینتەرنێتەوە قسە لەگەڵ ڕاژەکاری کۆمپانیا دەکات کە کۆپیی سەرەکی هەڵدەگرێت؛ بێ ئینتەرنێت لەوانەیە چاوەڕێ بکەیت یان هەر نەتوانیت فایلەکەت بکەیتەوە. ئەپی ناوخۆیی لە پێشدا: لاپتۆپەکەت کۆپیی سەرەکی هەڵدەگرێت؛ تیرێکی پچڕپچڕ دەچێت بۆ باکئەپ و هاوکاتکردنی ئارەزوومەندانە؛ بێ ئینتەرنێت کار دەکات و دەستبەجێ دەکرێتەوە و خۆت باکئەپ دروست دەکەیت. خوارەوە: حەوت ئامانجەکەی نەرمەواڵەی ناوخۆیی لە پێشدا (Ink & Switch، 2019): خێرا، چەند ئامێر، بێ ئینتەرنێت، هاوکاری، تەمەنی درێژ، تایبەتێتی، دەسەڵاتی بەکارهێنەر.
Where the main copy lives changes speed, privacy and who makes the backups.شوێنی کۆپیی سەرەکی خێرایی، و ئەوەی کێ دروست دەکات دەگۆڕێت.CêIT

Every button is also a command: CLI, JSON and MCPهەر دوگمەیەک فەرمانێکیشە: CLI، JSON و MCP

The apps can also be driven from a command line, through a JSON control channel (another program sends JSON messages) or through an MCP server, so that an AI agent can use them (MCP). The design idea: the engine holds every feature as a named command, and the windows are only one door to it. It is the thin-interface idea of Bazaar Cart's layers.

دەتوانرێن لە ڕێگەی نێوبەستەی هێڵی فەرمان، کەناڵێکی کۆنترۆڵی ( تر پەیامی دەنێرێت) یان MCP بەکار بهێنرێن، بۆ ئەوەی بتوانێت بەکاریان بهێنێت (پرۆتۆکۆڵی کۆنتێکستی مۆدێل). بیرۆکەی دیزاینەکە: بزوێنەرەکە هەموو تایبەتمەندییەک وەک فەرمانێکی ناودار هەڵدەگرێت و پەنجەرەکان تەنیا یەک دەرگان بۆی. ئەمە هەمان بیرۆکەی تەنکە لە چینەکانی Bazaar Cart.

Here is the idea in a few lines of Python for Bazaar Cart (prices are made up, in IQD): one command, two doors.

ئەمە بیرۆکەکەیە بە چەند دێڕێکی بۆ Bazaar Cart (نرخەکان دەستکردن، بە دیناری عێراقی): یەک فەرمان، دوو دەرگا.

cart_tool.pyPYTHONType in a terminalلە تێرمیناڵدا بینووسە
# cart_tool.py - one set of commands, two doors: the command line and JSON
import json
import math
import sys

PRICES = {"walnuts": 18_000, "sumac": 8_000, "figs": 12_000}  # IQD per kg (made up)


def round_up_250(amount):
    return math.ceil(amount / 250) * 250


def price(product, grams):
    """The command: R1, price per kg x grams / 1000, rounded up to 250 IQD."""
    return round_up_250(PRICES[product] * grams / 1000)


COMMANDS = {"price": price}  # the engine: every feature is a named command

if __name__ == "__main__":
    if sys.argv[1:] == ["--json"]:  # door 2: one JSON request per line
        for line in sys.stdin:
            request = json.loads(line)
            result = COMMANDS[request["command"]](**request["args"])
            print(json.dumps({"ok": True, "result": result}))
    else:  # door 1: the command line
        command, product, grams = sys.argv[1:4]
        print(COMMANDS[command](product, int(grams)))
Explain the codeڕوونکردنەوەی کۆدەکە
Lineدێڕی 13-15

The only command: rule R1, the price of one line, rounded up to 250 IQD.

تاکە فەرمان: یاسای ⁦R1⁩، نرخی یەک دێڕ، کە بەرەو سەرەوە بۆ ⁦250⁩ دینار خڕ دەکرێتەوە.

Lineدێڕی 18

The engine: a table of named commands. Every door looks the command up here.

بزوێنەرەکە: خشتەیەکی فەرمانی ناودار. هەموو دەرگایەک فەرمانەکە لێرە دەدۆزێتەوە.

Lineدێڕی 21-25

Door 2: another program sends one JSON request per line and gets a JSON answer.

دەرگای 2: تر لە هەر دێڕێکدا داواکارییەکی دەنێرێت و وەڵامێکی وەردەگرێتەوە.

Lineدێڕی 26-28

Door 1: a person or a script types the command in a terminal.

دەرگای 1: کەسێک یان فەرمانەکە لە تێرمیناڵدا دەنووسێت.

A central box, the engine, where every feature is a named command (price, open, export): one rule, one place. Four doors send arrows into it: buttons and windows, for people; the command line, for scripts and you in a terminal; a JSON channel, for other programs; an MCP server, for AI agents. Bottom: the windows are only one door; the same commands answer the CLI, JSON and AI agents, like a thin user-interface layer on top of the business layer.سندوقێکی ناوەڕاست، بزوێنەرەکە، کە هەر تایبەتمەندییەک تێیدا فەرمانێکی ناودارە (price، open، export): یەک یاسا، یەک شوێن. چوار دەرگا تیر دەنێرنە ناوی: دوگمە و پەنجەرەکان بۆ خەڵک؛ نێوبەستەی هێڵی فەرمان بۆ دەسنووسەکان و تۆ لە تێرمیناڵدا؛ کەناڵی JSON بۆ بەرنامەکانی تر؛ ڕاژەکاری MCP بۆ ئەیجێنتەکان. خوارەوە: پەنجەرەکان تەنیا یەک دەرگان؛ هەمان فەرمانەکان وەڵامی CLI، JSON و ئەیجێنتەکان دەدەنەوە، وەک چینێکی تەنکی نێوبەستەی بەکارهێنەر لەسەر چینی لۆجیکی کاروبار.
One engine of commands, four ways in.یەک بزوێنەری فەرمانەکان، چوار ڕێگای چوونەژوورەوە.CêIT

Open source, clean room and trademarksکۆدکراوە، دیزاینی ژووری پاک و نیشانە بازرگانییەکان

Most of ArtCraft's app repositories on GitHub use the Apache License 2.0, and PhotoCraft and DeckCraft let you choose MIT or Apache 2.0. Both are permissive licences: anyone may use, change, share and even sell the code if they keep the licence notice. Apache 2.0 also gives a patent licence, asks you to mark the files you changed, and gives no right to use the trademarks.

زۆربەی ArtCraft لە GitHub مۆڵەتی ⁦Apache 2.0⁩ بەکار دەهێنن و PhotoCraft و DeckCraft ڕێگەت پێدەدەن MIT یان ⁦Apache 2.0⁩ هەڵبژێریت. هەردووکیان مۆڵەتی ئاسانگیرن: هەر کەسێک دەتوانێت کۆدەکە بەکاربهێنێت، بیگۆڕێت، هاوبەشی بکات و تەنانەت بیفرۆشێت، بەو مەرجەی ئاگادارکردنەوەی مۆڵەتەکە بهێڵێتەوە. ⁦Apache 2.0⁩ هەروەها مۆڵەتی دەدات، داوا دەکات ئەو گۆڕیوتن نیشانە بکەیت و هیچ مافێکی بەکارهێنانی نادات.

PhotoCraft says it rebuilt Photoshop by clean-room design: from public specifications, such as Adobe's PSD file format, and from watching how the program behaves, never from Adobe's code. Its README adds that it is not affiliated with Adobe, and ArtCraft's own logos are not open source. Clean-room design helps against copying claims but not against patents, and a law expert noted that the claim has not been tested (Al Jazeera, 14 October 2026).

PhotoCraft دەڵێت Photoshopـی بە دروستکردۆتەوە: لە وەسفە گشتییەکان، وەک PSDـی Adobe و لە سەیرکردنی هەڵسوکەوتی ، هەرگیز لە کۆدی Adobe نا. READMEـەکەی هەروەها دەڵێت هیچ پەیوەندییەکی بە Adobeـەوە نییە و لۆگۆکانی خودی ArtCraft نین. یارمەتی دەدات دژی بانگەشەی لەبەرگرتنەوە، بەڵام دژی نا و پسپۆڕێکی یاسا ئاماژەی کرد کە ئەم بانگەشەیە هێشتا تاقی نەکراوەتەوە (Al Jazeera، ⁦14⁩ی تشرینی یەکەمی ⁦2026⁩).

The original, with closed code and public file formats, is studied by Team A, which looks at what it does, its specifications and how it behaves, and writes a description. A dashed wall stands between the teams: no code passes the wall, only the description. Team B, which never saw the original code and uses only the description, builds new code. Notes: a tick, helps against copying (copyright) claims; a cross, does not protect against patents; a cross, names and logos stay the owner's trademarks. Bottom: the classic example is Phoenix, which rebuilt the IBM PC BIOS this way in the 1980s.ئەسڵەکە، بە کۆدی داخراو و جۆرەکانی فایلی گشتی، تیمی A لێی دەکۆڵێتەوە: چی دەکات، وەسفە گشتییەکانی و چۆن هەڵسوکەوت دەکات و وەسفێک دەنووسێت. دیوارێکی پچڕپچڕ لە نێوان دوو تیمەکەدایە: هیچ کۆدێک بە دیوارەکەدا تێناپەڕێت، تەنیا وەسفەکە. تیمی B، کە هەرگیز کۆدی ئەسڵەکەی نەبینیوە و تەنیا وەسفەکە بەکار دەهێنێت، کۆدی نوێ دروست دەکات. تێبینییەکان: نیشانەی ڕاست، یارمەتی دەدات دژی بانگەشەی لەبەرگرتنەوە (مافی لەبەرگرتنەوە)؛ خاچ، لە بڕوانامەکانی داهێنان ناتپارێزێت؛ خاچ، ناو و لۆگۆکان نیشانەی بازرگانیی خاوەنەکەیان دەمێننەوە. خوارەوە: نموونەی کلاسیک Phoenixـە، کە لە ساڵانی 1980دا بەم شێوەیە BIOSـی IBM PCـی دروست کردەوە.
Clean-room design copies what a program does, never its code. ئەوە لەبەر دەگرێتەوە کە دەیکات، هەرگیز کۆدەکەی نا.CêIT

2026: why software companies were nervous⁦2026⁩: بۆچی کۆمپانیاکانی نەرمەواڵە نیگەران بوون

Adobe and Microsoft sell software that people use every day, mostly as subscriptions. In 2026 investors began to fear that AI tools could do part of that work, so that fewer people would pay. A share price shows what investors expect, not what customers do today.

Adobe و Microsoft دەفرۆشن کە خەڵک هەموو ڕۆژێک بەکاری دەهێنن، زۆربەی جار وەک . لە ⁦2026⁩دا وەبەرهێنەران ترسیان لێ نیشت کە ئامرازەکانی بتوانن بەشێک لەو کارە بکەن، بۆیە خەڵکی کەمتر پارە بدەن. نرخی پشک ئەوە پیشان دەدات کە وەبەرهێنەران چاوەڕێی دەکەن، نەک ئەوەی کڕیاران ئەمڕۆ دەیکەن.

  • 3 February: after Anthropic gave Claude Cowork, its agent for office work, new plugins (one for legal work), about $285 billion of share value was lost in one day across software, financial-services and asset-management companies (Bloomberg).
  • By 31 March: Microsoft's shares had fallen 23% in three months, their worst quarter since 2008, and Adobe's were down more than 30% for the year (CNBC).
  • 7 July: Bank of America rated Adobe 'Underperform', saying that generative AI makes creating content easier and brings cheaper, AI-native rivals.
  • 3ی شوبات: دوای ئەوەی Anthropic چەند نوێی بۆ Claude Cowork، بۆ کاری ئۆفیس، بڵاو کردەوە کە یەکێکیان بۆ کاری یاسایی بوو، لە یەک ڕۆژدا نزیکەی ⁦285⁩ ملیار دۆلار بەهای پشک لە کۆمپانیاکانی ، خزمەتگوزاریی دارایی و بەڕێوەبردنی سامان ڕۆیشت ⁦(Bloomberg)⁩.
  • تا ⁦31⁩ی ئازار: پشکی Microsoft لە سێ مانگدا ⁦23%⁩ دابەزیبوو، کە خراپترین چارەکی ساڵ بوو لە ⁦2008⁩ـەوە و پشکی Adobe ئەمساڵ زیاتر لە ⁦30%⁩ دابەزیبوو ⁦(CNBC)⁩.
  • 7ی تەمموز: Bank of America هەڵسەنگاندنی «Underperform»ـی (کەمتر لە بازاڕ) بە Adobe دا و گوتی دروستکردنی ناوەڕۆک ئاسانتر دەکات و ڕکابەری هەرزانتر دەهێنێت کە لە بنەڕەتەوە لەسەر دروستکراون.
A vertical timeline of 2026. 12 January: Anthropic releases Claude Cowork, an AI agent for office work (preview). 30 January: Cowork gets plugins, one of them for legal work. 3 February: about $285 billion of share value lost in one day (Bloomberg). 31 March: Microsoft down 23% in three months; Adobe down more than 30% this year (CNBC). 7 July: Bank of America rates Adobe Underperform, saying AI makes creating easier. 10 September: Adobe reports a record quarter, $6.76 billion, up 13%. 22 September: Anthropic releases Claude Opus 5.5. 7 to 12 October: the Craft Apps are in the news, the Office-like apps on 12 October. 13 October: Microsoft shares close at a record high (CNBC). November: the apps are still alpha or in development, and no report shows harm. Share-fall events are orange; records and the Craft Apps are green. Bottom: most of the fall came months before the Craft Apps existed.هێڵی کاتیی ستوونیی 2026. 12ی کانوونی دووەم: Anthropic ئەیجێنتی Claude Cowork بۆ کاری ئۆفیس بڵاو دەکاتەوە (تاقیکاری). 30ی کانوونی دووەم: Cowork پاشکۆکان وەردەگرێت، یەکێکیان بۆ کاری یاسایی. 3ی شوبات: نزیکەی 285 ملیار دۆلار بەهای پشک لە یەک ڕۆژدا ڕۆیشت (Bloomberg). 31ی ئازار: Microsoft لە سێ مانگدا 23% دابەزی؛ Adobe ئەمساڵ زیاتر لە 30% دابەزیوە (CNBC). 7ی تەمموز: Bank of America بە Adobe دەڵێت Underperform و دەڵێت ژیریی دەستکرد دروستکردن ئاسان دەکات. 10ی ئەیلوول: Adobe بەرزترین داهاتی چارەکێک ڕادەگەیەنێت، 6.76 ملیار دۆلار، 13% زیاتر. 22ی ئەیلوول: Anthropic مۆدێلی Claude Opus 5.5 بڵاو دەکاتەوە. 7 تا 12ی تشرینی یەکەم: Craft Apps دەچنە ناو هەواڵەکان، ئەپە هاوشێوەکانی Office لە 12ی تشرینی یەکەم. 13ی تشرینی یەکەم: پشکی Microsoft بە بەرزترین نرخی مێژووی دادەخرێت (CNBC). تشرینی دووەم: ئەپەکان هێشتا ئەلفان یان لە گەشەپێداندان و هیچ ڕاپۆرتێک زیان پیشان نادات. ڕووداوەکانی دابەزینی پشک پرتەقاڵین؛ بەرزترین ئاستەکان و Craft Apps سەوزن. خوارەوە: زۆربەی دابەزینەکە چەند مانگێک پێش بوونی Craft Apps ڕوویدا.
Dated events of 2026, each with its source.ڕووداوە ڕێکەوتدارەکانی ⁦2026⁩، هەر یەکێک بە سەرچاوەکەیەوە.CêIT

Microsoft's fall also had its own causes: investors doubted its huge AI spending and the slow uptake of its Copilot assistant (CNBC). Now look at the dates: most of the fall came months before the Craft Apps existed. It was a fear of AI tools in general, not a reaction to ArtCraft.

دابەزینی Microsoft هۆکاری تایبەتی خۆیشی هەبوو: وەبەرهێنەران گومانیان لە خەرجییە زۆرەکەی بۆ و بەکارهێنانی کەمی یاریدەدەری Copilot هەبوو ⁦(CNBC)⁩. ئێستا سەیری ڕێکەوتەکان بکە: زۆربەی دابەزینەکە چەند مانگێک پێش بوونی Craft Apps ڕوویدا. ترس بوو لە ئامرازەکانی بە گشتی، نەک کاردانەوە بەرامبەر ArtCraft.

What still protects them, and what is too early to knowچی هێشتا دەیانپارێزێت و چی هێشتا زووە بزانرێت

Free rivals are not new (see the picture below). Big products are protected by:

ڕکابەری بێبەرانبەر شتێکی نوێ نییە (بڕوانە وێنەکەی خوارەوە). بەرهەمە گەورەکان بەمانە دەپارێزرێن:

  • File formats: clients' PSD, DOCX and XLSX files must open exactly right.
  • Skills and plug-ins: years of training, and PhotoCraft runs no Photoshop plug-ins.
  • Cloud, contracts and support that companies pay for.
  • Strong sales and their own AI: Adobe's revenue hit a record $6.76 billion (+13%) in the quarter to 28 August 2026, and its 'AI-first' recurring revenue grew over 150% (Adobe).
  • : PSD، DOCX و XLSXـی کڕیاران دەبێت بە تەواوی دروست بکرێنەوە.
  • شارەزایی و : ساڵانێک ڕاهێنان و PhotoCraft هیچ Photoshop کار پێناکات.
  • هەور، گرێبەست و پشتگیری کە کۆمپانیاکان پارەیان بۆ دەدەن.
  • فرۆشی بەهێز و خۆیان: داهاتی Adobe لەو چارەکەی لە ⁦28⁩ی ئابی ⁦2026⁩ کۆتایی هات گەیشتە بەرزترین ئاست، ⁦6.76⁩ ملیار دۆلار ⁦(+13%)⁩ و داهاتی بەردەوامی بەرهەمە «⁦AI-first⁩»ـەکانی زیاتر لە ⁦150%⁩ گەشەی کرد ⁦(Adobe)⁩.
Everyday analogyنموونەیەک لە ژیانی ڕۆژانەوە

A new bakery opens next to a famous old one in the bazaar and gives bread away. The old bakery does not close the next day: it has regular customers and a taste people know. But if the new bread is just as good, it may have to lower its prices.

نانەواخانەیەکی نوێ لە تەنیشت نانەواخانەیەکی کۆن و بەناوبانگ لە بازاڕدا دەکرێتەوە و نان بە خۆڕایی دەدات. نانەواخانە کۆنەکە ڕۆژی دواتر دەرگای دانناخات: کڕیاری هەمیشەیی و تامێکی ناسراوی هەیە. بەڵام ئەگەر نانە نوێیەکە هەمان ئەوەندە باش بێت، لەوانەیە ناچار بێت نرخەکانی دابەزێنێت.

Where the analogy stops workingسنووری ئەم نموونەیە

Software is copied almost for free, so this can happen much faster.

بە نزیکەی بێ تێچوو لەبەر دەگیرێتەوە، بۆیە ئەمە دەتوانێت زۆر خێراتر ڕوو بدات.

A balance with a question mark at the pivot. Left pan, pressures: free copies built fast with AI, investors fear AI tools, AI may do part of the work, people tired of subscriptions. Right pan, protections: file formats and compatibility; skills, habits, plug-ins; cloud, teamwork, contracts, support; their own AI features. The beam is level. Below: in November 2026 it is too early to know which side will win; free rivals are not new, GIMP (1996) and LibreOffice (2011) did not replace Photoshop or Office.تەرازوویەک کە نیشانەی پرسیاری لە ناوەڕاستدایە. تای چەپ، فشارەکان: کۆپیی بێبەرانبەر کە بە خێرایی بە ژیریی دەستکرد دروستکراون، ترسی وەبەرهێنەران لە ئامرازەکانی ژیریی دەستکرد، لەوانەیە ژیریی دەستکرد بەشێک لە کارەکە بکات، خەڵکی بێزار لە ئابوونە. تای ڕاست، پارێزەرەکان: جۆرەکانی فایل و گونجان؛ شارەزایی، خوو، پاشکۆکان؛ هەور، کاری تیمی، گرێبەست، پشتگیری؛ تایبەتمەندییەکانی ژیریی دەستکردی خۆیان. تەرازووەکە ڕێکە. خوارەوە: لە تشرینی دووەمی 2026دا زووە بزانرێت کام لا دەباتەوە؛ ڕکابەری بێبەرانبەر نوێ نییە، GIMP (1996) و LibreOffice (2011) جێی Photoshop و Officeـیان نەگرتەوە.
Pressures and protections, balanced in November 2026.فشار و پارێزەرەکان، هاوسەنگ لە تشرینی دووەمی ⁦2026⁩دا.CêIT

What is new is speed: AI may help small teams copy features fast. Al Jazeera's analysis: 'less profitable', not obsolete. Will such apps replace traditional ones? In November 2026 it is too early to know: they are alpha, and no company results for the months after October 2026 exist yet.

ئەوەی نوێیە خێراییە: لەوانەیە یارمەتی تیمە بچووکەکان بدات تایبەتمەندییەکان بە خێرایی لەبەر بگرنەوە. شیکاریی Al Jazeera: «قازانجی کەمتر»، نەک لەناوچوون. ئایا وەک ئەمانە جێی دەگرێتەوە؟ لە تشرینی دووەمی ⁦2026⁩دا زووە بزانرێت: ئەلفان و هێشتا هیچ ئەنجامێکی کۆمپانیاکان بۆ مانگەکانی دوای تشرینی یەکەمی ⁦2026⁩ نییە.

Common mistakesهەڵە باوەکان

A coding mistake when you drive an app through JSON: writing the request with Python-style single quotes. JSON needs double quotes, so json.loads() stops the tool.

هەڵەیەکی کاتێک لە ڕێگەی بەکار دەهێنیت: نووسینی داواکارییەکە بە کۆتەیشنی تاکی وەک . کۆتەیشنی جووتی دەوێت، بۆیە json.loads() ئامرازەکە ڕادەگرێت.

Common mistake: single quotes in JSONهەڵەی باو: کۆتەیشنی تاک لە JSONـداBASHType in a terminalلە تێرمیناڵدا بینووسە
echo "{'command': 'price', 'args': {'product': 'figs', 'grams': 250}}" | python3.13 cart_tool.py --json
The fix: double quotesچارەسەرەکە: کۆتەیشنی جووتBASHType in a terminalلە تێرمیناڵدا بینووسە
echo '{"command": "price", "args": {"product": "figs", "grams": 250}}' | python3.13 cart_tool.py --json

Exercisesڕاهێنانەکان

Five blue cards joined by arrows, the questions to ask about a claim: 1, who says it? the company, a reporter, an analyst? 2, primary source? a README, a licence, a results report. 3, do the dates fit? a cause comes before its effect. 4, measured or guessed? numbers, tests, benchmarks. 5, other causes? what else happened at the same time? An arrow leads to three verdicts: TRUE, the evidence supports it; FALSE, the evidence contradicts it; NOT PROVEN, no good evidence either way yet. Bottom: a careful 'not proven' is better than a confident mistake; write the source and its date next to every number you repeat.پێنج کارتی شین کە بە تیر بەیەکەوە بەستراون، ئەو پرسیارانەی لەبارەی بانگەشەیەک دەکرێن: 1، کێ دەیڵێت؟ کۆمپانیاکە، ڕۆژنامەنووس یان شیکار؟ 2، سەرچاوەی سەرەکی؟ README، مۆڵەت، ڕاپۆرتی ئەنجام. 3، ڕێکەوتەکان دەگونجێن؟ هۆکار پێش ئەنجامەکەی دێت. 4، پێوراوە یان گومانە؟ ژمارە، تاقیکردنەوە، پێوەرەکانی کارایی. 5، هۆکاری تر؟ لە هەمان کاتدا چی تر ڕوویدا؟ تیرێک دەچێت بۆ سێ بڕیار: ڕاستە، بەڵگەکان پشتگیری دەکەن؛ هەڵەیە، بەڵگەکان پێچەوانەین؛ نەسەلمێنراوە، هێشتا بەڵگەی باش بۆ هیچ لایەک نییە. خوارەوە: گوتنی «نەسەلمێنراوە» بە وردبینی باشترە لە هەڵەیەکی پڕ لە متمانە؛ لە تەنیشت هەر ژمارەیەک کە دووپاتەی دەکەیتەوە سەرچاوە و ڕێکەوتی بنووسە.
Use the five questions in every exercise below.پێنج پرسیارەکە لە هەموو ڕاهێنانەکانی خوارەوەدا بەکاربهێنە.CêIT
ExerciseڕاهێنانJudge the claimsبڕیار لەسەر بانگەشەکان بدە

True, false or not proven? Give your evidence for each: (a) 'ArtCraft destroyed Adobe.' (b) 'Fear of AI tools pushed software shares down in 2026.' (c) 'The Craft Apps are open source.' (d) 'PhotoCraft can replace Photoshop for professional work today.' (e) 'Native Rust apps are always faster than Electron apps.' (f) 'ArtCraft's Office-like apps hurt Microsoft's share price.'

ڕاستە، هەڵەیە یان نەسەلمێنراوە؟ بۆ هەر یەکێک بەڵگەکەت بهێنەرەوە: ⁦(a)⁩ «ArtCraft Adobeـی لەناو برد.» ⁦(b)⁩ «ترس لە ئامرازەکانی لە ⁦2026⁩دا پشکی کۆمپانیاکانی دابەزاند.» ⁦(c)⁩ «Craft Apps .» ⁦(d)⁩ «PhotoCraft ئەمڕۆ دەتوانێت بۆ کاری پیشەیی جێی Photoshop بگرێتەوە.» ⁦(e)⁩ « کۆمپایلکراوی Rust هەمیشە لە Electron خێراترە.» ⁦(f)⁩ « Officeـی ArtCraft زیانیان بە نرخی پشکی Microsoft گەیاند.»

Show answerوەڵامەکە پیشان بدە
  • (a) False as stated, and any harm is not proven. On 10 September 2026 Adobe reported a record quarter ($6.76 billion, up 13%). Its share fall came mostly in February to June, before the apps existed, and no later results show an effect yet.
  • (b) True: Bloomberg (4 February) and CNBC (February and March) report it. It shows investors' fear, not proof that AI has replaced software.
  • (c) True: the licences on GitHub are Apache 2.0, and PhotoCraft and DeckCraft offer MIT or Apache 2.0; both are open-source licences.
  • (d) False: PhotoCraft's own README says it is not yet a replacement for daily professional work.
  • (e) Not proven, and false as a rule: no benchmarks are published, and speed depends on the program and its algorithm.
  • (f) Not proven, and the dates speak against it: Microsoft's shares closed at a record on 13 October 2026, one day after the Office-like apps were in the news; its earlier fall had other causes.
  • ⁦(a)⁩ هەڵەیە وەک گوتراوە و هەر زیانێکیش نەسەلمێنراوە. لە ⁦10⁩ی ئەیلوولی ⁦2026⁩دا Adobe بەرزترین داهاتی چارەکێکی ڕاگەیاند (⁦6.76⁩ ملیار دۆلار، ⁦13%⁩ زیاتر). دابەزینی پشکەکەی زۆربەی لە شوبات تا حوزەیران بوو، پێش بوونی و هیچ ئەنجامێکی دواتر هێشتا کاریگەرییەک پیشان نادات.
  • ⁦(b)⁩ ڕاستە: ⁦Bloomberg (4⁩ی شوبات) و CNBC (شوبات و ئازار) ڕاپۆرتیان کردووە. ترسی وەبەرهێنەران پیشان دەدات، نەک بەڵگە کە جێی گرتۆتەوە.
  • ⁦(c)⁩ ڕاستە: لە ⁦GitHub Apache 2.0⁩ـن و PhotoCraft و DeckCraft MIT یان ⁦Apache 2.0⁩ پێشکەش دەکەن؛ هەردووکیان .
  • ⁦(d)⁩ هەڵەیە: READMEـی خودی PhotoCraft دەڵێت هێشتا جێگرەوە نییە بۆ کاری پیشەیی ڕۆژانە.
  • ⁦(e)⁩ نەسەلمێنراوە و وەک یاسایەکی گشتی هەڵەیە: هیچ بڵاو نەکراوەتەوە و خێرایی پشت بە و دەبەستێت.
  • ⁦(f)⁩ نەسەلمێنراوە و ڕێکەوتەکان دژی دەوەستن: پشکی Microsoft لە ⁦13⁩ی تشرینی یەکەمی ⁦2026⁩دا بە بەرزترین نرخ داخرا، ڕۆژێک دوای ئەوەی Office چوونە ناو هەواڵەکان؛ دابەزینە پێشووەکەی هۆکاری تری هەبوو.
ExerciseڕاهێنانCheck 'native, not wrapped' yourselfخۆت «کۆمپایلکراو، نەک پێچراو» بپشکنە

How can you check, without trusting anyone, whether a desktop app carries its own browser engine? Plan a check, then compare with ours.

چۆن دەتوانیت، بێ ئەوەی متمانە بە کەس بکەیت، بپشکنیت ئایا سەر مێز خۆی لەگەڵ خۆیدا هەڵدەگرێت؟ پلانی پشکنینێک دابنێ، پاشان لەگەڵ ئەوەی ئێمە بەراوردی بکە.

Show answerوەڵامەکە پیشان بدە

Look inside the app. On a Mac: right-click it, choose Show Package Contents and open Contents/Frameworks. An Electron app has a folder named Electron Framework.framework; on any system, files ending in .asar (Electron's archive format) are a strong sign. We checked two apps on one Mac on 10 November 2026:

سەیری ناو بکە. لەسەر Mac: کلیکی ڕاستی لێ بکە، Show Package Contents هەڵبژێرە و Contents/Frameworks بکەرەوە. Electron فۆڵدەرێکی بە ناوی Electron Framework.framework هەیە؛ لەسەر هەر سیستەمێک، ئەو بە .asar کۆتاییان دێت (شێوازی ئەرشیفی Electron) نیشانەیەکی بەهێزن. ئێمە لە ⁦10⁩ی تشرینی دووەمی ⁦2026⁩دا دوو لەسەر یەک Mac پشکنی:

Measured on one Mac, 10 November 2026پێوراو لەسەر یەک Mac، ⁦10⁩ی تشرینی دووەمی ⁦2026⁩
AppBrowser engine inside? لە ناوەوە؟Size on diskقەبارە لەسەر دیسک
Visual Studio Code 1.145.2Yes: Electron Framework.framework, about 520 MBبەڵێ: Electron Framework.framework، نزیکەی ⁦520 MB⁩whole app about 1.6 GBهەموو نزیکەی ⁦1.6 GB⁩
DeckCraft 0.15.0No: one program that uses the system’s AppKit and Metalنەخێر: یەک کە AppKit و Metalـی سیستەم بەکار دەهێنێتwhole app about 237 MBهەموو نزیکەی ⁦237 MB⁩

So 'no Electron' is true for DeckCraft, and you can check it yourself. But native is not automatically small: DeckCraft's program holds two complete copies, one for Intel and one for Apple silicon (about 120 MB and 116 MB), and it ships licence files for 29 fonts. And size on disk says nothing about speed.

کەواتە «بێ Electron» بۆ DeckCraft ڕاستە و خۆت دەتوانیت بیپشکنیت. بەڵام کۆمپایلکراو خۆبەخۆ بچووک نییە: DeckCraft دوو کۆپیی تەواو هەڵدەگرێت، یەکێک بۆ Intel و یەکێک بۆ Apple silicon (نزیکەی ⁦120 MB⁩ و ⁦116 MB⁩) و مۆڵەتنامەی ⁦29⁩ لەگەڵ خۆیدا دەهێنێت. هەروەها قەبارە لەسەر دیسک هیچ لەبارەی خێرایی ناڵێت.

ExerciseڕاهێنانAlgorithm or language?ئەلگۆریزم یان زمان؟

Change the Python program to the sieve of Eratosthenes: start with all numbers marked prime and cross out the multiples of each prime. Time it. Is the new Python program faster or slower than the old C program?

بگۆڕە بۆ بێژینگی ئێراتۆستێنس: سەرەتا هەموو ژمارەکان بە سەرەتایی نیشانە بکە و پاشان چەندجارەکانی هەر ژمارەیەکی سەرەتایی بسڕەوە. کاتەکەی بپێوە. ئایا لە C خێراترە یان خاوتر؟

Show answerوەڵامەکە پیشان بدە
sieve.pyPYTHONType in a terminalلە تێرمیناڵدا بینووسە
# sieve.py - the same answer with a better algorithm (sieve of Eratosthenes)
def count_primes(limit):
    is_prime = [True] * limit
    is_prime[0] = is_prime[1] = False
    for n in range(2, int(limit ** 0.5) + 1):
        if is_prime[n]:
            for multiple in range(n * n, limit, n):
                is_prime[multiple] = False
    return sum(is_prime)

print(count_primes(1_000_000))
BASHType in a terminalلە تێرمیناڵدا بینووسە
/usr/bin/time -p python3.13 sieve.py
/usr/bin/time -l python3.13 sieve.py 2>&1 | grep "maximum resident"

Same answer, 78,498, in about 0.1 seconds: roughly as fast as the C program with the slow method, and about 50 to 90 times faster than the old Python program. It uses more memory (about 21 MB), because it keeps a list of a million Boolean values (True or False). A better algorithm beat a faster language; choose the algorithm first, then measure.

هەمان وەڵام، ⁦78,498⁩، لە نزیکەی ⁦0.1⁩ چرکەدا: نزیکەی هێندەی C بە ڕێگا خاوەکە خێرایە و نزیکەی ⁦50⁩ تا ⁦90⁩ هێندە لە خێراترە. زیاتر بەکار دەهێنێت (نزیکەی ⁦21 MB⁩)، چونکە لیستێکی یەک ملیۆن بەهای (True یان False) هەڵدەگرێت. باشتر زمانێکی خێراتری بردەوە؛ سەرەتا هەڵبژێرە، پاشان بیپێوە.

ExerciseڕاهێنانMake the JSON door safeدەرگای JSON سەلامەت بکە

Change cart_tool.py so that a bad request gets an error answer instead of crashing the tool. Test it with four requests in a file: single quotes, a good request, an unknown product (honey) and an unknown command (delete_all).

cart_tool.py بگۆڕە بۆ ئەوەی داواکارییەکی خراپ وەڵامێکی هەڵە وەربگرێت لە جیاتی ئەوەی ئامرازەکە بوەستێنێت. بە چوار داواکاری لە فایلێکدا تاقیی بکەرەوە: کۆتەیشنی تاک، داواکارییەکی باش، کاڵایەکی نەناسراو (هەنگوین) و فەرمانێکی نەناسراو (delete_all).

Show answerوەڵامەکە پیشان بدە
cart_tool.py (version 2): the JSON door⁦cart_tool.py⁩ (وەشانی ⁦2⁩): دەرگای JSONPYTHONType in a terminalلە تێرمیناڵدا بینووسە
COMMANDS = {"price": price}  # the engine: every feature is a named command

if __name__ == "__main__":
    if sys.argv[1:] == ["--json"]:  # door 2: one JSON request per line
        for line in sys.stdin:
            try:
                request = json.loads(line)
                result = COMMANDS[request["command"]](**request["args"])
                print(json.dumps({"ok": True, "result": result}))
            except (ValueError, KeyError, TypeError) as error:
                print(json.dumps({"ok": False, "error": str(error)}))
    else:  # door 1: the command line
        command, product, grams = sys.argv[1:4]
        print(COMMANDS[command](product, int(grams)))

Every bad request now gets an answer with "ok": false, and the tool keeps working. The unknown command delete_all is refused because the engine runs only the commands in its table. That is a safety rule too: an AI agent can ask only for what the program offers.

ئێستا هەموو داواکارییەکی خراپ وەڵامێک بە "ok": false وەردەگرێت و ئامرازەکە بەردەوام کار دەکات. فەرمانی نەناسراوی delete_all ڕەت دەکرێتەوە، چونکە بزوێنەرەکە تەنیا ئەو فەرمانانە کار پێدەکات کە لە خشتەکەیدان. ئەمە یاسایەکی سەلامەتیشە: تەنیا دەتوانێت داوای ئەوە بکات کە پێشکەشی دەکات.

ExerciseڕاهێنانReview an AI-written pull requestبە داواکاریی بەیەککردنێکی نووسراو بە ژیریی دەستکرددا بچۆرەوە

An AI agent sends Bazaar Cart a pull request that adds rule R5 (5% off goods totals of 100,000 IQD or more, rounded up to 250). Write five checks you will make before you merge it.

بۆ Bazaar Cart دەنێرێت کە یاسای ⁦R5⁩ زیاد دەکات (⁦5%⁩ داشکان بۆ کۆی کاڵا لە ⁦100,000⁩ دینار یان زیاتر، بەرەو سەرەوە بۆ ⁦250⁩ خڕ دەکرێتەوە). پێنج پشکنین بنووسە کە پێش دەیکەیت.

Show answerوەڵامەکە پیشان بدە
  • Compare the code with R5 word by word: 5%, from 100,000 IQD, rounded up to 250.
  • Run all old tests and add new ones at the boundary: 99,750 and 100,000 IQD (test cases).
  • Read every line until you can explain it to a teammate.
  • Check any new dependency's licence, and that no secret or customer data was added.
  • Plan the release: a new feature means version 1.1.0 and a changelog line (versions).
  • کۆدەکە وشە بە وشە لەگەڵ ⁦R5⁩ بەراورد بکە: ⁦5%⁩، لە ⁦100,000⁩ دینارەوە، بەرەو سەرەوە بۆ ⁦250⁩ خڕ دەکرێتەوە.
  • هەموو تاقیکردنەوە کۆنەکان کار پێبکە و تاقیکردنەوەی نوێ لە سنووردا زیاد بکە: ⁦99,750⁩ و ⁦100,000⁩ دینار (هەڵبژاردنی حاڵەتەکان).
  • هەموو دێڕێک بخوێنەرەوە تا بتوانیت بۆ هاوتیمێکی ڕوون بکەیتەوە.
  • هەر نوێ بپشکنە و دڵنیا بە هیچ نهێنییەک یان کڕیار زیاد نەکراوە.
  • پلانی بڵاوکردنەوە دابنێ: تایبەتمەندییەکی نوێ واتە ⁦1.1.0⁩ و دێڕێک لە (وەشانەکان).
ExerciseڕاهێنانSwitch or not?بگۆڕێت یان نا؟

Soran asks whether his shop should move from Word and Excel to WordCraft and GridCraft to save money. What would you tell him in November 2026?

سۆران دەپرسێت ئایا دووکانەکەی بۆ پاشەکەوتکردنی پارە لە Word و Excelـەوە بگوازێتەوە بۆ WordCraft و GridCraft. لە تشرینی دووەمی ⁦2026⁩دا چی پێ دەڵێیت؟

Show answerوەڵامەکە پیشان بدە

Not yet for daily work: both apps are still in development. He can test them on copies first: open his real .docx and .xlsx files from suppliers and the bank, check Kurdish right-to-left text, fonts and printing, plan his own backups, and ask who will help when something breaks. Keep the current tools until the free ones pass these tests; free is not the same as no cost.

هێشتا بۆ کاری ڕۆژانە نا: هەردوو هێشتا لە گەشەپێداندان. دەتوانێت سەرەتا لەسەر کۆپییەکان تاقییان بکاتەوە: .docx و .xlsx کە لە دابینکەران و بانکەوە هاتوون بکاتەوە، دەقی کوردیی ، و چاپکردن بپشکنێت، پلانی خۆی دابنێت و بپرسێت کاتێک شتێک تێک دەچێت کێ یارمەتی دەدات. ئامرازە ئێستاییەکان بهێڵێتەوە تا بێبەرانبەرەکان لەم تاقیکردنەوانە دەردەچن؛ بێبەرانبەر هەمان «بێ تێچوو» نییە.

Summaryکورتە

  • Craft Apps (November 2026): 16 free, open-source Rust apps; four early alpha, twelve in development, versions 0.x.
  • Reportedly built by one engineer with Claude Opus 5.5 agents; their rules (tests, layers, reviewed pull requests, licences) are ordinary software engineering.
  • Rust compiles to machine code and to WebAssembly; the apps draw their own windows instead of carrying a browser engine like Electron.
  • Compiled, local-first apps can be fast and light, but only measurements show it, and the algorithm often matters most.
  • Every feature as a named command lets people, scripts, other programs and AI agents (through MCP) use one engine.
  • Permissive licences (MIT, Apache 2.0) allow reuse with a notice; clean-room design avoids copying code but not patents; logos stay trademarks.
  • Software shares, Adobe's and Microsoft's included, fell in early 2026 on fears of AI tools in general, months before the Craft Apps. No evidence shows that ArtCraft harmed either company; whether such apps will replace traditional ones is too early to know.
  • Craft Apps (تشرینی دووەمی ⁦2026): 16⁩ بێبەرانبەر و Rust؛ چواریان سەرەتایی، دوازدەیان لە گەشەپێداندا، ⁦0.x⁩.
  • بەپێی ڕاپۆرتەکان یەک ئەندازیار بە ⁦Claude Opus 5.5⁩ دروستی کردوون؛ یاساکانیان (تاقیکردنەوە، ، کە پێیاندا چوونەتەوە، ) ئاسایین.
  • Rust بۆ و بۆ کۆمپایل دەکرێت؛ پەنجەرەکانی خۆیان دەکێشن لە جیاتی ئەوەی وەک Electron هەڵبگرن.
  • کۆمپایلکراو و ناوخۆیی لە پێشدا دەتوانێت خێرا و سووک بێت، بەڵام تەنیا پێوانە ئەمە پیشان دەدات و زۆرجار گرنگترینە.
  • هەر تایبەتمەندییەک وەک فەرمانێکی ناودار ڕێگە دەدات خەڵک، ، تر و (لە ڕێگەی MCP) یەک بزوێنەر بەکاربهێنن.
  • (MIT، ⁦Apache 2.0⁩) ڕێگە بە دووپاتە بەکارهێنان دەدەن بە هێشتنەوەی ئاگادارکردنەوە؛ خۆی لە لەبەرگرتنەوەی کۆد دەپارێزێت، بەڵام نەک لە ؛ لۆگۆکان دەمێننەوە.
  • پشکی کۆمپانیاکانی ، هی Adobe و Microsoftـیش، لە سەرەتای ⁦2026⁩دا بەهۆی ترس لە ئامرازەکانی بە گشتی دابەزین، چەند مانگێک پێش Craft Apps. هیچ بەڵگەیەک پیشان نادات کە ArtCraft زیانی بە هیچ یەک لەو دوو کۆمپانیایە گەیاندبێت؛ هێشتا زووە بزانرێت ئایا وەک ئەمانە جێی دەگرێتەوە.

Next: the course project puts these rules to work: tests, layers, reviews and a real 1.0.0 release.

دواتر: پڕۆژەی کۆرس ئەم یاسایانە دەخاتە کار: تاقیکردنەوە، ، پێداچوونەوە و ڕاستەقینەی ⁦1.0.0⁩.

Lesson complete!وانەکە تەواو بوو!

Well done. Review the key terms, then go on to the next lesson.دەستخۆش. سەیرێکی زاراوە سەرەکییەکان بکەرەوە، پاشان بڕۆ بۆ وانەی داهاتوو.

Key termsزاراوە سەرەکییەکان

Next lessonوانەی داهاتوو

Course Project: Ship Bazaar Cart 1.0.0پڕۆژەی خولەکە: ⁦Bazaar Cart 1.0.0⁩ بڵاو بکەرەوە

References and further readingسەرچاوەکان و خوێندنەوەی زیاتر

This lecture is a first edition. The Kurdish text and terminology are waiting for review by Kurdish IT teachers.ئەم وانەیە چاپی یەکەمە. دەقە کوردییەکە و زاراوەکانی چاوەڕێی پێداچوونەوەی مامۆستایانی کوردی بواری IT ن.

Appearanceڕووکار

Modeدۆخ
Styleشێواز
Coloursڕەنگەکان

Saved on this device only.تەنیا لەسەر ئەم ئامێرە پاشەکەوت دەکرێت.