dom ด้้้้้ ‏

News addict.Bullshiteur artistique,Utopian but still believe in humans.

[Hugo Caron AI @hugocaron_IA]

Si ce thread t’a été utile, abonne-toi @hugocaron_IA

pour plus de contenu sur l’IA, le
business et la tech.

[Girişimci Hisler @girisimcihisler]

İnternet hiçbir şeyi unutmaz diyorlar. Yanılıyorlar. Ona unutturmayı bilmeniz gerek.

Geçen gün birinin Google Takeout verilerini inceledik. 2001 yılından beri, tam 185 GB veri birikmiş. Konum geçmişinden yıllar önce sildiğini sandığı ses kayıtlarına kadar her şey orada.

İnternetteki varlığınız saatli bir bomba gibi beklemesin. "Dijital Ayak İzinizi" sıfırlamak sandığınızdan daha kolay. Bu bir amme hizmetidir. Kaydedin, uygulayın.

1. Büyük Veriyle Yüzleşin (Google Takeout)
Önce Google Takeout'a girin ve hakkınızda ne bildiklerini indirin. Gördüklerinize şaşıracaksınız. İyi haber şu: Google'a düzenli aralıklarla (3-6 ayda bir) mail atıp "Hakkımdaki verileri silin" deme hakkınız kanunen var. Buna "Hayır" diyemezler.

2. Dijital Mezarlığı Temizleyin
Yıllar önce açıp unuttuğunuz o forum üyelikleri, utanç verici eski tweetler... Hepsini bulmak için "JustDeleteMe" veya "AccountKiller" araçlarını kullanın.
Mailiniz bir yerlerden sızdı mı? "HaveIBeenPwned" üzerinden kontrol edin. Sızıntı varsa şifrenizi değil, dijital kimliğinizi değiştirme vakti gelmiştir.

3. Görsel Hafızayı Silin
Adınızı arattığınızda çıkanlar buzdağının görünen kısmı. "TinEye" veya Google Görseller ile yüzünüzü ters arama yapın. İstemediğiniz fotoğrafların olduğu sitelere içerik kaldırma formu gönderin. Üşenmeyin.

4. Veri Tüccarlarına (Data Broker) Dur Deyin
Spokeo veya Whitepages gibi siteler verilerinizi toplayıp satar. Bunlardan manuel olarak "opt-out" (sistemden çıkma) talebi oluşturun. Sizin veriniz üzerinden para kazanmalarına izin vermeyin.

5. Telefon Numarası Tuzağı
Numaranız sizi dijital dünyada "siz" yapan en güçlü koddur. Mümkün olan her yerden numaranızı kaldırın. SMS doğrulaması yerine uygulama tabanlı 2FA (Google Authenticator vb.) kullanın. Numaranızı vermeyin, ProtonMail gibi servislerle ikincil (alias) kimlikler oluşturun.

6. Araç Çantası (Setup)
Sıradan tarayıcıları bırakın.
• Tarayıcı: Brave veya Firefox.
• Arama: Startpage.
• Kalkan: uBlock Origin ve Privacy Badger eklentileri.
• Hafıza: Her yer için farklı şifre üreten Bitwarden.

Fotoğraflarınızın deepfake malzemesi, verilerinizin reklam ürünü olmasını istemiyorsanız "Dijital Hijyen" artık tercih değil, zorunluluktur.

[Tarık Tolunay 🏛️ Kentçizer @tariktolunay]

Bu sıradan bir şehir çizimi değil.

Paris. 🗼

10 ay. Tek kişi.
15 milyar piksel.
Sıfır yapay zekâ.

İçine gir. 🧐
https://t.co/eE9fwoz1Sc

[Mario Alexis @ALEXIS_74B]

@KeylinAcosta Aquí tienes la versión completa de la reentrada atmosférica de Artemis 1 sin cortes, grabada directamente en la memoria de la cámara

[sui ☄️ @birdabo]

we got an astronaut posting on 𝕏 from a spacecraft that’s about halfway to the moon.

what a time to be alive.

[Déborah @dvorahfr]

Quality mode of Grok Imagine is insane

[GnosisWolf @GnosisWolf]

She’s a keeper. I hope she makes it. 🙏

[Russian Market @runews]

🇺🇸 If the United States and Israel continue striking the Bushehr nuclear power plant in Iran, radiation could threaten the entire Middle East.

Dubai and the United Arab Emirates would be at risk of radioactive contamination.

[Nav Toor @heynavtoor]

🚨BREAKING: Claude has a hidden feature called "Learning Mode."

It turns Claude into a personal tutor that teaches you anything step-by-step.

Here's how to use it 👇

[Jouhatsu | AI Influence Operator @Jouhatsu_ai]

C'est DINGUE.

J'ai uploadé mes relevés bancaires à Claude Opus 4.6.

Il a analysé une année complète de transactions, séparé revenus et dépenses, et tout catégorisé.

20 minutes plus tard, la comptabilité était terminée.
Environ 11 000€ de travail comptable réglé en une session.

Voici le prompt exact que j'ai utilisé + 5 autres prompts bonus :

[ Ajoutez en signet 🔖pour ne pas perdre ! ]

[Okara @askOkara]

Today we're introducing the world's first AI CMO.

Enter your website and it deploys a team of agents to help you get traffic and users.

Try it now at https://t.co/KbAE6FNgzE

[Kshitij Mishra | AI & Tech @DAIEvolutionHub]

Someone put free TV channels from around the world into a single playlist.

Just load the M3U file into VLC or an IPTV player and watch.

Open-source.
Continuously updated.

https://t.co/JsEzhLzkdv

[Eloi | UGC IA 🧨 @EloiLJF]

BREAKING : Anthropic vient de lancer des formations gratuites pour maîtriser l'IA avec certificats.

Sans frais.
Sans liste d'attente.
Sans BS.

Voici 10 cours qui complètent votre formation en IA :

[ Ajoutez en signet 🔖pour ne pas perdre ! ]

[code w/Guillaume 💻 @guillaume_rygn]

très bon repo à connaitre 👇

[Supersocks @iamsupersocks]

Le mec qui a créé Claude Code (@bcherny) vient de montrer comment son équipe dresse l’IA.

Un fichier. CLAUDE.md. Tu le poses à la racine de ton projet. Dedans : les erreurs passées, les conventions, les règles. Claude le lit à chaque session.

Résultat : l’agent s’améliore sans que tu retouches une ligne de code. Chaque bug corrigé devient une règle permanente.
Boris Cherny utilise ça tous les jours chez Anthropic. Je vous mets son template ici.

Prêt à copier/coller et à adapter à votre guise :

### 1. Plan Mode Default
- Enter plan mode for ANY non-trivial task (3+ steps or architectural decisions)
- If something goes sideways, STOP and re-plan immediately — don't keep pushing
- Use plan mode for verification steps, not just building
- Write detailed specs upfront to reduce ambiguity

### 2. Subagent Strategy
- Use subagents liberally to keep main context window clean
- Offload research, exploration, and parallel analysis to subagents
- For complex problems, throw more compute at it via subagents
- One task per subagent for focused execution

### 3. Self-Improvement Loop
- After ANY correction from the user: update `tasks/lessons. md` with the pattern
- Write rules for yourself that prevent the same mistake
- Ruthlessly iterate on these lessons until mistake rate drops
- Review lessons at session start for relevant project

### 4. Verification Before Done
- Never mark a task complete without proving it works
- Diff behavior between main and your changes when relevant
- Ask yourself: "Would a staff engineer approve this?"
- Run tests, check logs, demonstrate correctness

### 5. Demand Elegance (Balanced)
- For non-trivial changes: pause and ask "is there a more elegant way?"
- If a fix feels hacky: "Knowing everything I know now, implement the elegant solution"
- Skip this for simple, obvious fixes — don't over-engineer
- Challenge your own work before presenting it

### 6. Autonomous Bug Fixing
- When given a bug report: just fix it. Don't ask for hand-holding
- Point at logs, errors, failing tests — then resolve them
- Zero context switching required from the user
- Go fix failing CI tests without being told how

## Task Management

1. **Plan First**: Write plan to `tasks/todo.md` with checkable items
2. **Verify Plan**: Check in before starting implementation
3. **Track Progress**: Mark items complete as you go
4. **Explain Changes**: High-level summary at each step
5. **Document Results**: Add review section to `tasks/todo. md`
6. **Capture Lessons**: Update `tasks/lessons. md` after corrections

## Core Principles

- **Simplicity First**: Make every change as simple as possible. Impact minimal code.
- **No Laziness**: Find root causes. No temporary fixes. Senior developer standards.

[Shanaka Anslem Perera ⚡ @shanaka86]

BREAKING: A Chinese AI startup called MizarVision is publishing high-resolution satellite imagery of every US military base, every carrier strike group, every F-22 deployment, every THAAD battery, and every Patriot missile position in the Middle East. Labelled. Geolocated. AI-annotated. Updated in near-realtime. Shared by PLA-linked accounts and Chinese state media to an audience of billions.

The first major release came on 20 February, eight days before Operation Epic Fury began. MizarVision published images showing US aircraft transfers to Ovda Airbase in southern Israel, fighter deployments across Saudi Arabia and Qatar, and naval buildups in the Arabian Sea. By 1 March, the releases had expanded to include detailed imagery of bases in Jordan, Kuwait, Bahrain, and the UAE, with AI labelling identifying specific aircraft types, air defence configurations, and troop concentrations. One release catalogued approximately 2,500 individual US military assets across the region.

The imagery comes from two sources. The first is China’s Jilin-1 satellite constellation, a network of over 100 commercial Earth observation satellites operated by Chang Guang Satellite Technology, whose data is used by the PLA. A majority of Jilin-1 satellites are dedicated to regional imaging with sub-metre resolution, capable of identifying individual aircraft on tarmacs and distinguishing between THAAD and Patriot battery configurations from orbit. The second source is commercially available Western satellite data from providers like Maxar and Airbus, which MizarVision aggregates, processes through proprietary AI models for automatic target recognition, and republishes with military-grade labelling that transforms raw imagery into actionable intelligence products.

The Pentagon has downplayed the releases as “open-source.” This framing misses the point entirely.

The value of MizarVision’s output is not the raw satellite image. Any government can purchase commercial satellite passes. The value is the AI processing layer that converts terabytes of imagery into labelled, searchable, cross-referenced intelligence products at a speed and scale that previously required the resources of a national intelligence agency. MizarVision is democratising military surveillance and publishing the output on social media where Iran’s 31 autonomous IRGC provincial commands can access it from a mobile phone.

No direct evidence confirms classified data transmission from Beijing to Tehran. But the distinction between “classified” and “publicly shared AI-processed satellite intelligence identifying every US military asset in the Middle East by type, location, and configuration” is a distinction without a meaningful difference to a provincial IRGC commander selecting his next target.

The strategic implications extend far beyond this conflict. In the 2022 Ukraine war, Maxar’s commercial satellite imagery aided Kyiv by exposing Russian deployments. The West celebrated it as the democratisation of intelligence. China has now executed the identical playbook in reverse: a nominally commercial firm, with documented PLA data-sharing arrangements, publishing intelligence products that expose American deployments during an active war. The precedent is set. Commercial satellite intelligence is now a weapon of great-power competition deployed through AI startups with plausible commercial deniability.

MizarVision has fewer than 200 employees. Its AI models run on commercially available hardware. Its satellite data comes from constellations any nation can build. And it has just demonstrated the capability to map every US military asset across an entire theatre of war and publish the results on the open internet before the first bomb falls.

The next war will not begin with a missile launch. It will begin with an AI model labelling every target from orbit.

Full analysis - https://t.co/eMrt5qYYst

[Bruno Gaccio @GaccioB]

Excellent article. Long et documenté.
A lire si ce qui se passe en Iran vous intéresse un peu, sinon la télé suffit: tu sauras le superficiel Trumpien .