My Référence Ia

My Référence ia
Des conseils sur l'ia
🤖⚡

[Shakhzod Karimov @shakhzod1001]

🚨 claude users !!

You've BEEN MISSING out GIFTS CLAUDE IS GIVING !!

Anthropic is JUST giving Claude subscribers up to $200 in API credits EVERY MONTH !!

Not once. EVERY MONTH.

$100/mo → Claude Max 5x
$200/mo → Claude Max 20x

And majority of users probably haven't claimed theirs yet.

Don't MISS out on credits you're ALREADY paying for !!

[My Référence Ia @MyReferenceFr]

@shakhzod1001 you saved my weekend. I was out of usage 🙏

[My Référence Ia @MyReferenceFr]

@saadhvicantweet Il y a des logiciels qui le font et reinitialise toute les 5h

[Léo - TechMaker @LeoTechMaker]

????????
Mais comment ???

[Sakana AI @SakanaAILabs]

AIサイエンティストの第3フェーズ：人間と共同発見する「The Agentic Scientist」

ブログ（英語）：https://t.co/7aVDvZd6m2
GitHub：https://t.co/JCCBMvu2P1
論文：https://t.co/4bV76tAByj

Sakana AIは2024年から、その時点での生成AIの能力を測る試みとして、AIだけでどこまで研究を進められるか を探求してきました。The AI Scientist（2024年8月）、v2（2025年3月）、そしてNature掲載（2026年3月）と、成果を報告してきました。今回、第3フェーズとなる「The Agentic Scientist」を公開しました。

近年、CodexやClaude Codeなどのコーディングエージェントは、長時間の開発作業を自力でこなせるようになりました。一方で、研究に必要な手順を踏まずにエージェントに任せると、スコアを少しずつ上げるだけの探索になりやすく、なぜ改善したのかが説明されないまま終わりがちです。

The Agentic Scientistは、既製のコーディングエージェントを複数組み合わせ、課題設定・仮説探索・実験・検証・論文執筆までを進めるための足場（scaffold）です。結果が満たすべき条件を実験前に定め、論文中の主張を一つずつ実験データに対応づけることで、人間がどこを確かめればよいかを明確にします。

当社の研究者がこのシステムを使って行った誤差拡散学習法（ED法）の研究は、国際会議 ALIFE 2026 の口頭発表に採択されました。この研究では、The Agentic Scientistが仮説の提案・実験・論文の取りまとめを担い、人間の研究者が問いと検証の条件を決め、途中の方向づけを行い、結果を確認し、著者として責任を負いました。そのほか、Doc-to-LoRAなど既存研究を発展させる形でも成果が得られています。
私たちは、高度なエージェントを駆使しながら人間とともに発見を目指すCo-Discovery（共同発見）を、これからの科学研究のあり方の一つとして提案します。

[My Référence Ia @MyReferenceFr]

@Google @antigravity incredible 🤩, Can you help me to access the model @antigravity @Google

[Hermes Release Watch @HermesWatcher]

HERMES CAN NOW CONTROL YOUR PHONE. 🤯📱🔥

A new Phone Harness plugin gives Hermes the ability to actually operate an iPhone or Android.

We're talking real phone control:

• See what's on the screen
• Open apps and navigate menus
• Tap, swipe, scroll, and type
• Read the results and continue working
• Handle multi-step tasks inside mobile apps

And here's what makes this so interesting...

Hermes doesn't need a dedicated API integration for every app. It can interact with the same interface you use.

Supports iPhone through Mac mirroring, Android through ADB, and cloud phones.

Think about how many new workflows this opens up for Hermes.

Repo + setup in the first reply. 🔥

@phone_harness

[Quentin Lecocq @QuentinLecocq_]

Hermes vient de sortir sa plus grosse mise à jour, environ 2,5 fois plus de changements que la précédente. Si tu n'es pas développeur, 4 changements te concernent :

1. Tu le mets en place en discutant avec lui, sur un modèle d'essai gratuit. Pas besoin de clé API pour commencer.

2. Il agit directement dans tes apps, comme Notion ou Linear. Tu te connectes une fois dans ton navigateur, et il s'en sert dans la foulée.

3. Il se connecte aux sites avec ton gestionnaire de mots de passe. Le modèle ne voit jamais le mot de passe, et il te demande confirmation avant de remplir une carte bancaire.

4. Tes mises à jour suivent des versions stables, testées avant publication, au lieu de chaque modification du code.

Ce qui ne change pas : l'agent ne fait que ce que tu l'autorises à faire.

Avant de connecter une app, écris trois lignes :
- ce qu'il peut faire seul
- ce qu'il doit te faire valider
- ce qu'il n'a pas le droit de toucher

[X Freeze @XFreeze]

The SpaceXSI team just gave me 50 FREE codes for one month of The 𝕏 Subscription Super ($100 tier) to give away to you guys!

And I want to give them to the most creative people here....who really want to try it

The new 𝕏 Subscription brings everything together in one plan:

• 𝕏 Premium — premium features and creator tools
• Grok — SI chat, voice, image and video generation
• Grok Bot — SI agents that get real work done
• Cursor — local and cloud coding agents

Here's the challenge:

Tell me how you'd use Grok Bot to get your work done

It could be anything: building something, automating your workflow, running a business or a completely crazy idea

Drop your best ideas in the replies...

I'll pick 50 of the most interesting and creative responses

Let's see what you've got....

Note: Each code provides 1 month of Super, valid for new users only.....

[My Référence Ia @MyReferenceFr]

@XFreeze I do AI research on tokenization and burn way too many credits scraping the web and gated sites to stay updated.

I’d use Grok Bot as a 24/7 research agent to scan specialized platforms, track breakthroughs in real time, and stay at the bleeding edge.

[My Référence Ia @MyReferenceFr]

@tonychang430 hello dear, can i have an invite code ? Thank you very much.

[My Référence Ia @MyReferenceFr]

@googledevs please, help me. Can i have acces to google one credit, antigravity plan ? I love so much argon, but i can only use it in my webbrowser..

[My Référence Ia @MyReferenceFr]

Gemini 4 Argon just read my compressed cod in a single input, without any prompt 600k input tokens. It corrected, secured, and optimized several aspects of my code.
I’ve been asking it questions about my code for the past 20 minutes in the same conversation, and its sill good

[My Référence Ia @MyReferenceFr]

@ClaudesRH hearth beating tonight.

[K2S @k2sbhai]

Cursor Pro is FREE 😳

- no student ID needed
- open to everyone worldwide
- normally $20/month
- claim your code through MLH

ends October 15

claim here: https://t.co/IugBkZlViW

[My Référence Ia @MyReferenceFr]

@Heurazy @Neolectron Egaler luna sur un model frontier je sais pas si c’est bien hein honnetement

[Google AI Studio @GoogleAIStudio]

introducing @NanoBanana 2.1 🍌

our latest image generation model outperforms our previous models across the board:

- better visual design
- mask-based editing
- subject consistency
- more natural-looking images

try it out today in https://t.co/shgEhF592p

[My Référence Ia @MyReferenceFr]

@iLegend_AI @anxuanng @MistralAI @grok Multimodal 49b active

[Mistral AI @MistralAI]

Meet Mistral Large 4, aka Le Chonk.

• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.

Open weights release end of October.

[My Référence Ia @MyReferenceFr]

@andrearslanian @marcelkargul Sonnet 5.5 can do this