نسخهٔ اپل اپ CCloud را ساختم.
نسخهٔ آزمایشی iPhone، iPad و Apple TV روی TestFlight و نسخه Mac رو میتونید از گیتهاب دانلود کنید.
TestFlight: https://t.co/TE3TIAAuJ2
GitHub: https://t.co/kPwBcftzDD
#movie#ccloud
apple already built a way to swap siri for claude, gemini, or chatgpt. it's sitting in the ios 27 beta code. settings panel, app store section, all of it. they just turned it off and said nothing at wwdc.
#iOS27#AI
built a scroll animation into Maliyar that collapses the monthly calendar to weekly when you scroll up, so all your installments for that day actually fit on screen. small thing, but makes a real difference.
https://t.co/M05DvrPfvr #maliyr#buildinpublic
before going solo i would've just built it. in a team someone asks the hard questions. when you're alone, you either build that muscle or you ship a mess.
#buildinpublic#indiedev#solofounder
then i sat down and actually thought about it. adding it would change what the app is. Maliyar is an installment tracker, not a financial hub. so i didn't build it.
@cursor_ai just announced Origin, a version control platform built for AI agents.
22.6 commits per second in a single repo. hundreds of thousands of clones and pushes per hour.
no pricing, no early access, fall 2026 is the only confirmed window.
#ai#ai_tools
caveman does the other end. makes the model answer short instead of padded. real dollar saving was about $1 in a session. the actual win was faster cleaner answers.
https://t.co/qhuIy6KnuM
kept hitting token limits mid task, at work and on my own projects. AI stops, says come back later. found 2 open source tools that cut usage from opposite ends 🧵
rtk sits between your terminal and the model. compresses noisy command output (git, tests, grep) before it hits context. across ~2000 commands it cut about 52% of my input tokens.
https://t.co/oix2VXhKPI
anthropic says claude wrote 80%+ of their production code last month. one engineer hasn't coded in 5 months.
i'm an iOS dev. i shipped an iOS+Android app solo anyway.
the shift isn't coming. it's already here.
#ClaudeCode#AIEngineering
anthropic just released claude fable 5. stripe said it did a 50M-line codebase migration in one day. that's a 2-month job for a full team. testing it this week.
#Claude#AI
went back to my project and found exactly the part that had this problem. not frustrating honestly. more like something blurry finally came into focus. that's the best part of learning while building.
#buildinginpublic#AIEngineering#indiedev
been building with LLMs for months thinking "AI knows everything, just ask it." learned about RAG this week and realized a whole part of my app's architecture was wrong because of that assumption.
RAG is simple in concept: LLMs only know what they were trained on. your user's private data, anything that changed after the training cutoff, anything that was never public — the model has no idea. RAG bridges that gap without retraining the model.