Arthur
51 posts
Arthur
@ArthurSaada
10x Entreprenor — 1x Software engineer — building an app to manage your social media
Paris, France 가입일 Mayıs 2012
136 팔로잉16 팔로워
Today we introduce OASIS 1.
The smart ring built for private dictation. Whisper to write. Touch to edit.
A first step beyond the keyboard toward a world where your intent follows you across every device.
Order at oasisdevices.com first batch is limited.
English
i feel like i should gatekeep this but...
i think consumers are about to start prioritizing apps that have good connectors into AI assistants
mainly bc of the new siri update coming soon 👀
here is what i did this month to "future proof" my app to stay ahead
English
It took me 52 days from the idea to make an app to actually getting published to App Store
For an average X launch – 52 days is eternity
It's not about number of days
For me it was an interesting journey where I learned a ton before the first user arrives:
- no coding background
- no shortcuts with Rork/x10
- my first paywall (Superwall)
- integrating revenue tracking (RevenueCat)
- integrating product analytics (PostHog)
- deliberate design
- crafting an onboarding flow
- deciding on MVP features
Next: distribution, updates, $$$
Pasha@pasha_barbashin
I've scoured the App Store and found 30 niches packed with $50K+ apps Day 1 of building a profitable iOS app - zooming out to spot the real opportunities - hunting for something that gets me fired up - sizing up the competition
English
@pasha_barbashin Thanks ! Will check this way
Is it working ? How much ARR r you making now ?
English
@ArthurSaada Hey Arthur
I just followed this logic covered by Jake. RevenueCat for analytics, funnels, tracking. Superwall for experiments.
Jake Mor@jakemor
To those curious it’s not a sample size thing (cc @jackfriks) it holds true for all apps. See AppFigures data below: One hypothesis is RC is more popular for apps starting out. Younger apps = less revenue. Less time to grow. This isn’t true though - if we look at apps launched only since Jan 1, 2026 Superwall avg is $3.1k (n = 896) vs RevenueCat’s $960 (n = 16k) For apps launched after Jan 1 doing over $10k/mo - RC has 65 apps compared to SW having 15. 0.4% of apps using RC pass $10k/mo in under 6 months. (avg = $46k) 1.7% of apps using SW pass $10k/mo in under 6 months. (avg = $52k) @HHaandr and I looked into this about a year ago and the happy ending is that apps using us both actually do the best. Only 233 apps launched this year using us both, and the avg is $4.8k The best founders don’t build anything in house unless it gives them a competitive advantage. Doesn’t really matter what the cost is because they know these tools pay for themselves. They use SW for paywalls and AB testing. They use RC revenue analytics and subscription management 💥
English
At first sight I think it’s stupid to ask Claude in Slack instead of Claude code …
But after some research I understand it’s the first step from Anthropic toward agentic système and Agent product like Hermes/openclaw ..
what are your thoughts ?
Claude@claudeai
Introducing Claude Tag, a new way for teams to work with Claude. In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.
English
Arthur 리트윗함
We’ve designed and built our first AI chip: Jalapeño.
Designed from the ground up by OpenAI and brought to production with @Broadcom, Jalapeño is purpose-built for the LLM workloads powering ChatGPT, Codex, the API, and future agentic products.
Chips are foundational to the AI economy. Building our own expands our full-stack platform from products to models to infrastructure, and will help us scale intelligence, serve more people, and expand access to AI.
English
Nvidia CEO right now
OpenAI@OpenAI
We’ve designed and built our first AI chip: Jalapeño. Designed from the ground up by OpenAI and brought to production with @Broadcom, Jalapeño is purpose-built for the LLM workloads powering ChatGPT, Codex, the API, and future agentic products. Chips are foundational to the AI economy. Building our own expands our full-stack platform from products to models to infrastructure, and will help us scale intelligence, serve more people, and expand access to AI.
English
@bitcoindudebro @EngMoElgaraihy Yes exactly… on open router you can have 200token/s (not talking about fraction of the cost)
English
@EngMoElgaraihy @ArthurSaada Your point here is not really about the speed but about the hardware efficiency. That's a fair point but the fact remains that 21 tokens/s is very slow for working speed in coding projects. Still an interesting direction though.
English
فريق Unsloth المرعب قاموا بخطوة جنونية وأخذوا العملاق الصيني GLM 5.2 وضغطوه بتقنية متطرفة جدا تسمى 1-bit وحولوه لصيغة GGUF الخفيفة
المفاجأة؟ قاموا بتشغيله محلياً (Locally) بالكامل على جهاز Mac Studio وبدون أي اتصال بالإنترنت أو سيرفرات خارجية وبسرعة ممتازة بلغت 21 توكن في الثانية
ولم يكتفوا بذلك بل وضعوا هذا الموديل المحلي المضغوط في مواجهة حية ومباشرة ضد أعتى وأقوى سيرفرات العالم المدفوعة والسحابية Claude 4.8 Opus و GPT-5.5
الصدمة التي أذهلت المطورين في الكومنتات هي أن الموديل المحلي وقف نداً لند أمام سيرفرات بمليارات الدولارات وقدم إجابات ذكية ودقيقة جدا تضاهي العمالقة
الفائز الحقيقي اليوم ليس مجرد موديل بل هو مفهوم التشغيل المحلي؛ من الآن تصبح بياناتك آمنة 100% في جهازك وتكلفة فواتير الـ APIs الخاصة بك صفر دولار وبذكاء ينافس أقوى شركات أمريكا
لا تنسوا الإعجاب والمشاركة ليصل المنشور لكل مهووس بالتقنية والذكاء الاصطناعي !
Unsloth AI@UnslothAI
1-bit GLM-5.2 GGUF vs. Claude 4.8 Opus vs. GPT-5.5 We gave 3 models the same prompt and compared one-shot outputs. The 1-bit GLM-5.2 GGUF ran locally on a Mac Studio M3 Ultra with 256GB RAM at ~21.6 tok/s. Which output do you like best? GGUF: huggingface.co/unsloth/GLM-5.…
العربية
@DevonRJames @EngMoElgaraihy Sure! That’s why I will wait more years before do local
English
@ArthurSaada @EngMoElgaraihy it’s the slowest it’ll ever be. hardware will only get better, compression approaches will only get better.
English
Great rebranding of le chat from @MistralAI.
Vibe is a much stronger name. The interface seems to have improved as well. Good job
English
Tung Tung Tung Sahur is here!!
Day 14 of building GTA 6 with a loop of AI agents.
Today's progress:
- Reloads in. mag drops, new one goes in, old one on the ground.
- Shot effects landed.
- Full of Tung Tung Tung Sahur walking around the city now
We have to actually learn Unreal. if we don't, our prompts stay vague and the agents give vague results.
The second we need them to generate blueprints, animation blueprints, any of that, you have to know what to ask for.
The more specific i get, the better the output.
Tomorrow: aiming + all the full animations on the character.
Ziwen@ziwenxu_
Two days ago the loop built me a gun that couldn't shoot. Today I pulled the trigger and the whole street ran. Day 13 of building GTA 6 with a loop of AI agents. Remember the gun that wouldn't fire? and the crowd walking around arms-out like scarecrows? both gone today. Today's progress: - shooting works now. trigger pulls, round actually leaves the barrel. - the NPC walks better now. - then the part I wasn't ready for. I aimed at one of them and pulled the trigger, and the whole street panicked. they scattered and ran from me, terrified. first time the world's ever reacted to anything I did in it. next up: hit effects, gunshot sounds, reload.
English