Massive Kritik von Forschern: Anthropic ändert Richtlinie zur Entwicklung anderer KI-Modelle

Anthropic hat eine umstrittene Sicherheitsvorgabe für Fable 5 zurückgenommen, nachdem sie bei Forschern auf erhebliche Kritik gestoßen war.

ComputerBase

New Claude Fable 5 - fping complete rewrite as library from clang to ziglang in one 5 hour session:
https://github.com/zaxified/zig-fping

#fable5 #fable #claude #zig

GitHub - zaxified/zig-fping: fping rewritten in Zig — reusable ICMP probing library + fping-compatible CLI. Pure Zig, no libc, built for 10k+ target monitoring.

fping rewritten in Zig — reusable ICMP probing library + fping-compatible CLI. Pure Zig, no libc, built for 10k+ target monitoring. - zaxified/zig-fping

GitHub

Pohrál jsem si s #Fable5. Funguje dobře, 5h okno jsem v rámci #Claude Pro za 435 Kč vyčerpal za 45 min. (cca 500k tokenů, 10 % z týdenního limitu). Jednoduchým propočtem je v rámci Claude Pro při čtyřtýdenním měsíci k dispozici cca 20M tokenů.

Nicméně se zdá, že tento nový model už Anthropic nehodlá tolik dotovat, a brzy bude vyžadovat platbu dle využití. Těch 20M tokenů bude stát cca 25 600 Kč. A na "full time" by Fable vyšel na 127 500 Kč. Vše bez daně.

Párty končí.

Claude Fable 5を1日使ってみて - Qiita

はじめに 本日Anthropic一般公開したモデル「Fable 5」を、リリース翌日から丸一日、Claude Code でのコーディング・開発作業を中心に触ってみました。本記事はその所感を述べたものです。 モデル概要 まあ読んでいる時点で知らないことはないと思いますが前...

Qiita
Anthropic Claude Fable 5: Quellcode in mobiler App deutet weitere Funktionen an

Hinweise in einer aktuellen iOS-Version der Claude-App deuten auf mehrere geplante, aktuell aber noch inaktive Funktionen hin.

ComputerBase

Today's Challenge:

Stop Engineer A from using #Claude model B (#fable5) because he thinks it accelerates him, but rather it just hinders him.

👀 Claude Fable 5 e Mythos 5: Anthropic lancia lo stesso modello in due versioni — una pubblica e con filtri di sicurezza, l'altra senza restrizioni per i partner verificati del programma Glasswing
https://gomoot.com/claude-fable-5-e-mythos-5-anthropic-separa-la-potenza-dallaccesso/

#AI #Claude #fable5 #Mythos

Tre minuti. Due giorni. Il problema con l’AI che va troppo veloce. - Luca Mastronicola

Devo confessare una cosa: per due giorni ho litigato con un'AI. Non nel senso filosofico. Nel senso che continuavo a dirle "fixa questo bug" e lei continuava a non farcela. Il problema era tecnico ma banale: una landing page per i miei clienti, con un modulo che doveva allegare file e inviarli via email. Roba

Luca Mastronicola - Grafica & Web

Bright and early this morning I popped onto @BBC5Live's Wake Up to Money programme to chat to @seanfarrington about Anthropic's new AI model Claude Fable 5, which launched last night 🎙️

Is it safe to use, since it's technically a Mythos-class AI model, and there have been lots of cybersecurity concerns?

Also, this is kind of a full-circle milestone moment for me, as over 3 years, from 2017 to 2020, I would be on early shifts with @bbcbusiness transcribing interviews with guests for the BBC Business Live Blog (until the blog was retired), and now I've appeared on the show. Thanks for having me on! Younger me would be proud⭐😊

Listen from to find out more 🎧: https://www.bbc.co.uk/programmes/m002xgbc

#AI #Claude #Anthropic #Mythos #Fable5 #LLMs #generativeAI #AIagents #cybersecurity #bioweapons #technews #technology #analysis